Academia.eduAcademia.edu

Outline

Title
Abstract
Figures
Introduction
Planning with Hypotheses
Finding Explanations as a Planning Problem
Overall Quantitative Analysis
Quantitative Analysis
Discussion of Related Work
Conclusion
Summary of Contributions
Questions for Future Work
References
All Topics
Computer Science
Artificial Intelligence

Robot task planning and explanation in open and uncertain worlds

Profile image of Jeremy L WyattJeremy L Wyatt

Abstract

A long-standing goal of AI is to enable robots to plan in the face of uncertain and incomplete information, and to handle task failure intelligently. This paper shows how to achieve this. There are two central ideas. The first idea is to organize the robot's knowledge into three layers: instance knowledge at the bottom, commonsense knowledge above that, and diagnostic knowledge on top. Knowledge in a layer above can be used to modify knowledge in the layer(s) below. The second idea is that the robot should represent not just how its actions change the world, but also what it knows or believes. There are two types of knowledge effects the robot's actions can have: epistemic effects (I believe X because I saw it) and assumptions (I'll assume X to be true). By combining the knowledge layers with the models of knowledge effects, we can simultaneously solve several problems in robotics: (i) task planning and execution under uncertainty; (ii) task planning and execution in open worlds; (iii) explaining task failure; (iv) verifying those explanations. The paper describes how the ideas are implemented in a three-layer architecture on a mobile robot platform. The robot implementation was evaluated in five different experiments on object search, mapping, and room categorization.

Figures (22)
Figure 1: (Left) The instance-default-diagnostic (IDD) knowledge schema, showing action models (yellow), general knowledge (cream), instance knowledge about the current state (brown), and predicted state (blue). (Right) Example state information in the concrete instantiation of the IDD schema used in this paper. The blue box represents possible instance knowledge assumed to be true by the planner in order to achieve the robot’s goal (find a magazine). The right panel summarizes the robot’s belief state at the start of the running example (see Figure 3).
Figure 1: (Left) The instance-default-diagnostic (IDD) knowledge schema, showing action models (yellow), general knowledge (cream), instance knowledge about the current state (brown), and predicted state (blue). (Right) Example state information in the concrete instantiation of the IDD schema used in this paper. The blue box represents possible instance knowledge assumed to be true by the planner in order to achieve the robot’s goal (find a magazine). The right panel summarizes the robot’s belief state at the start of the running example (see Figure 3).
Figure 2: The overall architecture shows how the belief-state maintenance and the switching planner are structured and linked to the basic compe- tences for sensing and action. The implementation of this architecture is described in Section 6.
Figure 2: The overall architecture shows how the belief-state maintenance and the switching planner are structured and linked to the basic compe- tences for sensing and action. The implementation of this architecture is described in Section 6.
Figure 3: (Left) This shows the initial map state when Dora is switched on. Details are given in the text. (Right) This shows the plan generatec for the goal “find a magazine” from the initial state. Light blue entries are assumptive actions, green denotes the current action, white denotes ar unexecuted action, and orange denotes a planned call to the decision-theoretic planner. The probabilities of success for each step are taken, by the planner, from the relational map.
Figure 3: (Left) This shows the initial map state when Dora is switched on. Details are given in the text. (Right) This shows the plan generatec for the goal “find a magazine” from the initial state. Light blue entries are assumptive actions, green denotes the current action, white denotes ar unexecuted action, and orange denotes a planned call to the decision-theoretic planner. The probabilities of success for each step are taken, by the planner, from the relational map.
Figure 4: (Left) The map state, three minutes into a run, illustrates the view-planning problem handled by the POMDP planner. The purple tetrahedra are the view cones. (Right) The corresponding plan execution status; grey actions represent completed actions.
Figure 4: (Left) The map state, three minutes into a run, illustrates the view-planning problem handled by the POMDP planner. The purple tetrahedra are the view cones. (Right) The corresponding plan execution status; grey actions represent completed actions.
Figure 5: (Left) The robot in the office, at the time of failure. The view cones searched are coloured blue. The robot is in the centre of the room. (Right) The plan to verify the explanation of the failure, showing the three assumptions that need to be made.
Figure 5: (Left) The robot in the office, at the time of failure. The view cones searched are coloured blue. The robot is in the centre of the room. (Right) The plan to verify the explanation of the failure, showing the three assumptions that need to be made.
Action
Action
Figure 6: This shows the structure of the chain graph model, compiled from the relational model. The vertices represent random variables. The edges represent the directed and undirected probabilistic relationships between the random variables. The textured vertices indicate observations that correspond to sensed evidence.
Figure 6: This shows the structure of the chain graph model, compiled from the relational model. The vertices represent random variables. The edges represent the directed and undirected probabilistic relationships between the random variables. The textured vertices indicate observations that correspond to sensed evidence.
Algorithm 1 Continual switching planner
Algorithm 1 Continual switching planner
Figure 7: A visualization of the action dependencies of the plan from Figure 3. Assumptions are coloured blue, and physical actions are colourec grey. Arrows indicate that the first action has an effect which the second action depends on via preconditions or conditional effects.
Figure 7: A visualization of the action dependencies of the plan from Figure 3. Assumptions are coloured blue, and physical actions are colourec grey. Arrows indicate that the first action has an effect which the second action depends on via preconditions or conditional effects.
Figure 9: A continual planning process consisting of three planning cycles. Assumptions are coloured in blue, and executed actions are colourec in grey. The state resulting from execution of the last action in a plan (e.g. €o,1) is the initial state for the next planning cycle.
Figure 9: A continual planning process consisting of three planning cycles. Assumptions are coloured in blue, and executed actions are colourec in grey. The state resulting from execution of the last action in a plan (e.g. €o,1) is the initial state for the next planning cycle.
Figure 11: The action sequence (bottom row) enforced on the planner via the action-ordering variable M, so as to reproduce the executed portions of three, partially executed, plans (top three rows). Each observation action 0; requires that the immediately previous physical action resulted in the corresponding observed states 5;.
Figure 11: The action sequence (bottom row) enforced on the planner via the action-ordering variable M, so as to reproduce the executed portions of three, partially executed, plans (top three rows). Each observation action 0; requires that the immediately previous physical action resulted in the corresponding observed states 5;.
Figure 13: A place map with several places and three detected doors shown as red dots. Colours on circular discs indicate the probability of room categories as in a pie chart. Here green is corridor, red is kitchen and blue is office. Grey circles denote place-holders in unexplored space.
Figure 13: A place map with several places and three detected doors shown as red dots. Colours on circular discs indicate the probability of room categories as in a pie chart. Here green is corridor, red is kitchen and blue is office. Grey circles denote place-holders in unexplored space.
Figure 14: Creating view-point actions for object search. (Left) The prior probability distribution for an object in a room (purple), based on the locations of likely supporting surfaces (green). (Right) Examples of view cones with probabilities.
Figure 14: Creating view-point actions for object search. (Left) The prior probability distribution for an object in a room (purple), based on the locations of likely supporting surfaces (green). (Right) Examples of view cones with probabilities.
Figure 16: The robot’s maps at the end of the runs described in Sections 7.2 and 7.4. The offices are located on the left and at the top left. Three- dimensional obstacles are shown in green. The pie charts show room category probabilities. The category colours are as follows: purple for office, yellow for corridor and orange for meetingroom. Places are numbered, corresponding to the course of action plots in Figures 17, 18 and 20.
Figure 16: The robot’s maps at the end of the runs described in Sections 7.2 and 7.4. The offices are located on the left and at the top left. Three- dimensional obstacles are shown in green. The pie charts show room category probabilities. The category colours are as follows: purple for office, yellow for corridor and orange for meetingroom. Places are numbered, corresponding to the course of action plots in Figures 17, 18 and 20.
Figure 17: The sequence of the robot’s beliefs (top) and actions (bottom) during an exploration run. The top plot shows the room category distribution after each action. The blue line gives the entropy of the room category distributions, averaged over all rooms found. The numbers at the top refer to the robot’s location, corresponding to the annotations in Figure 16(a). The bottom plot shows the action sequence, indicating execution durations. The action type is colour-coded. Blanks indicate that the robot was planning for that period.
Figure 17: The sequence of the robot’s beliefs (top) and actions (bottom) during an exploration run. The top plot shows the room category distribution after each action. The blue line gives the entropy of the room category distributions, averaged over all rooms found. The numbers at the top refer to the robot’s location, corresponding to the annotations in Figure 16(a). The bottom plot shows the action sequence, indicating execution durations. The action type is colour-coded. Blanks indicate that the robot was planning for that period.
Figure 18: Course of action and the robot’s belief about the categories of rooms in an example categorization run, initiated at the end of the exploration run in Figure 17 that resulted in the map in Figure 16(a). Note the monotonic decrease in the robot’s uncertainty. The uncertainty about room category drops significantly after each dialogue (timestamps 323, 571, and 836 in Fig- ure 18). Figure 19 shows the plan structure: the assumptive actions cope with the uncertainty in room categories, by supposing a particular set of categories is true. Those assumptions are instance assumptions, and thus each assump- tion’s probability is taken from the relational map. The assumptions are also used to hypothesize people in each room who can be asked about the category of that room. Thus, it can be seen how the assumptive actions are used to reason about uncertain worlds with a classical planner. The dialogue actions are chosen when uncertainty is high. This is typically the case when only the laser is used for room categorization. TY..1..42.24£ 2.2... 4%2° 22 On Sf a 2. a. i 2 To ae yy. WT. .14 ato. 2 gh
Figure 18: Course of action and the robot’s belief about the categories of rooms in an example categorization run, initiated at the end of the exploration run in Figure 17 that resulted in the map in Figure 16(a). Note the monotonic decrease in the robot’s uncertainty. The uncertainty about room category drops significantly after each dialogue (timestamps 323, 571, and 836 in Fig- ure 18). Figure 19 shows the plan structure: the assumptive actions cope with the uncertainty in room categories, by supposing a particular set of categories is true. Those assumptions are instance assumptions, and thus each assump- tion’s probability is taken from the relational map. The assumptions are also used to hypothesize people in each room who can be asked about the category of that room. Thus, it can be seen how the assumptive actions are used to reason about uncertain worlds with a classical planner. The dialogue actions are chosen when uncertainty is high. This is typically the case when only the laser is used for room categorization. TY..1..42.24£ 2.2... 4%2° 22 On Sf a 2. a. i 2 To ae yy. WT. .14 ato. 2 gh
Figure 19: The first few steps of the initial plan during the categorization run depicted in Figure 18. Note the sequence of assumptive actions, the probabilities of which correspond to the final room category distributions at the end of the exploration run in Figure 17. The movement (yellow). sensing (red), and dialogue (red) actions have probability 1.00. The likelihood of plan success is the product of the probabilities; 0.27 is therefore the likelihood of the plan succeeding, according to the current belief state.
Figure 19: The first few steps of the initial plan during the categorization run depicted in Figure 18. Note the sequence of assumptive actions, the probabilities of which correspond to the final room category distributions at the end of the exploration run in Figure 17. The movement (yellow). sensing (red), and dialogue (red) actions have probability 1.00. The likelihood of plan success is the product of the probabilities; 0.27 is therefore the likelihood of the plan succeeding, according to the current belief state.
Figure 20: Course of actions in an example object-search run.
Figure 20: Course of actions in an example object-search run.
Table 1: Explanation runs. The first three runs (2, 9, and 14) listed correspond to runs in which the robot did not find the object because o perceptual errors. Runs el-e7 are a specific explain test condition, in which the object was deliberately hidden (condition C;,). We also report th number of actions the system carried out, the number of explanations generated, the most probable explanation, and the time f¢ it took to generat the explanations. The explanations are detailed on the right.
Table 1: Explanation runs. The first three runs (2, 9, and 14) listed correspond to runs in which the robot did not find the object because o perceptual errors. Runs el-e7 are a specific explain test condition, in which the object was deliberately hidden (condition C;,). We also report th number of actions the system carried out, the number of explanations generated, the most probable explanation, and the time f¢ it took to generat the explanations. The explanations are detailed on the right.

Related papers

Towards An Architecture for Representation, Reasoning and Learning in Human-Robot Collaboration
Mohan Sridharan

2016

Robots collaborating with humans need to represent knowledge, reason, and learn, at the sensorimotor level and the cognitive level. This paper summarizes the capabilities of an architecture that combines the comple- mentary strengths of declarative programming, proba- bilistic graphical models, and reinforcement learning, to represent, reason with, and learn from, qualitative and quantitative descriptions of incomplete domain knowledge and uncertainty. Representation and reasoning is based on two tightly-coupled domain representations at different resolutions. For any given task, the coarse- resolution symbolic domain representation is translated to an Answer Set Prolog program, which is solved to provide a tentative plan of abstract actions, and to explain unexpected outcomes. Each abstract action is implemented by translating the relevant subset of the corresponding fine-resolution probabilistic representation to a partially observable Markov decision process (POMDP). Any high pro...

View PDFchevron_right
Open-World Reasoning for Service Robots
Justin Hart

Proceedings of the International Conference on Automated Planning and Scheduling

A service robot accepting verbal commands from a human operator is likely to encounter requests that reference objects not currently represented in its knowledge base. In domestic or office settings, the construction of a complete knowledge base would be cumbersome and unlikely to succeed in most real-world deployments. The world that such a robot operates in is thus “open” in the sense that some objects that it must act on in the real world are not described in its internal representation. However, when an operator gives a command referencing an object that the robot has not yet observed (and thus not incorporated into its knowledge base), we can think of the object as being hypothetical to the robot. This paper presents a novel method for closing the robot’s world model for planning purposes by introducing hypothetical objects into the robot’s knowledge base, reasoning about these hypothetical objects, and acting on these hypotheses in the real world. We use our implementation of ...

View PDFchevron_right
A Refinement-Based Architecture for Knowledge Representation and Reasoning in Robotics
Jeremy L Wyatt

ABSTRACTThis paper describes an architecture that combines the complementary strengths of probabilistic graphical models and declarative programming to enable robots to represent and reason with logic-based and probabilistic descriptions of uncertainty and domain knowledge. An action language is extended to support non-boolean fluents and non-deterministic causal laws. This action language is used to describe tightly-coupled transition diagrams at two levels of granularity, refining a coarse-resolution transition diagram of the domain to obtain a fine-resolution transition diagram. The coarse-resolution system description, and a history that includes (prioritized) defaults, are translated into an Answer Set Prolog (ASP) program. For any given goal, inference in the ASP program provides a plan of abstract actions. To implement each such abstract action probabilistically, the part of the fine-resolution transition diagram relevant to this action is identified, and a probabilistic representation of the uncertainty in sensing and actuation is included and used to construct a partially observable Markov decision process (POMDP). The policy obtained by solving the POMDP is invoked repeatedly to implement the abstract action as a sequence of concrete actions, with the corresponding observations being recorded in the coarse-resolution history and used for subsequent reasoning. The architecture is evaluated in simulation and on a mobile robot moving objects in an indoor domain, to show that it supports reasoning with violation of defaults, noisy observations and unreliable actions, in complex domains.

View PDFchevron_right
Mixed Logical and Probabilistic Reasoning for Planning and Explanation Generation in Robotics
Mohan Sridharan

ArXiv, 2015

Robots assisting humans in complex domains have to represent knowledge and reason at both the sensorimotor level and the social level. The architecture described in this paper couples the non-monotonic logical reasoning capabilities of a declarative language with probabilistic belief revision, enabling robots to represent and reason with qualitative and quantitative descriptions of knowledge and degrees of belief. Specifically, incomplete domain knowledge, including information that holds in all but a few exceptional situations, is represented as a Answer Set Prolog (ASP) program. The answer set obtained by solving this program is used for inference, planning, and for jointly explaining (a) unexpected action outcomes due to exogenous actions and (b) partial scene descriptions extracted from sensor input. For any given task, each action in the plan contained in the answer set is executed probabilistically. The subset of the domain relevant to the action is identified automatically, a...

View PDFchevron_right
Towards an Explanation Generation System for Robots: Analysis and Recommendations
Ben Meadows

Robotics

A fundamental challenge in robotics is to reason with incomplete domain knowledge to explain unexpected observations and partial descriptions extracted from sensor observations. Existing explanation generation systems draw on ideas that can be mapped to a multidimensional space of system characteristics, defined by distinctions, such as how they represent knowledge and if and how they reason with heuristic guidance. Instances in this multidimensional space corresponding to existing systems do not support all of the desired explanation generation capabilities for robots. We seek to address this limitation by thoroughly understanding the range of explanation generation capabilities and the interplay between the distinctions that characterize them. Towards this objective, this paper first specifies three fundamental distinctions that can be used to characterize many existing explanation generation systems. We explore and understand the effects of these distinctions by comparing the capabilities of two systems that differ substantially along these axes, using execution scenarios involving a robot waiter assisting in seating people and delivering orders in a restaurant. The second part of the paper uses this study to argue that the desired explanation generation capabilities corresponding to these three distinctions can mostly be achieved by exploiting the complementary strengths of the two systems that were explored. This is followed by a discussion of the capabilities related to other major distinctions to provide detailed recommendations for developing an explanation generation system for robots.

View PDFchevron_right
Answer me this: Constructing Disambiguation Queries for Explanation Generation in Robotics
Mohan Sridharan

2021 IEEE International Conference on Development and Learning (ICDL), 2021

Robot assisting humans in complex domains can operate more effectively if they can describe their decisions and beliefs. Such transparency is difficult to achieve in integrated robot systems that include methods for learning from data and reasoning with incomplete commonsense domain knowledge. The architecture described in this paper seeks to address the associated challenges by building on a baseline system that supports non-monotonic logical reasoning with incomplete commonsense domain knowledge, data-driven learning from a limited set of examples, and inductive learning of previously unknown axioms governing domain dynamics. In the context of enabling a simulated robot to provide on-demand, relational descriptions of its decisions and beliefs, we introduce an interactive system that automatically traces beliefs, and constructs and poses queries to solicit human input and reduce ambiguity in the robot's beliefs. We present results of evaluation in scene understanding and planning tasks to demonstrate the abilities of our architecture.

View PDFchevron_right
Integrating Robot Task Planner with Common-sense Knowledge Base to Improve the Efficiency of Planning
Rossi Setchi

Procedia Computer Science, 2013

This paper presents a developed approach for intelligently generating symbolic plans by mobile robots acting in domestic environments, such as offices and houses. The significance of the approach lies in developing a new framework that consists of the new modeling of high-level robot actions and then their integration with common-sense knowledge in order to support a robotic task planner. This framework will enable interactions between the task planner and the semantic knowledge base directly. By using common-sense domain knowledge, the task planner will take into consideration the properties and relations of objects and places in its environment, before creating semantically related actions that will represent a plan. This plan will accomplish the user order. The robot task planner will use the available domain knowledge to check the next related actions to the current one and the action's conditions met will be chosen. Then the robot will use the immediately available knowledge information to check whether the plan outcomes are met or violated.

View PDFchevron_right
Towards a Theory of Explanations for Human–Robot Collaboration
Ben Meadows

KI - Künstliche Intelligenz

This paper makes two contributions towards enabling a robot to provide explanatory descriptions of its decisions, the underlying knowledge and beliefs, and the experiences that informed these beliefs. First, we present a theory of explanations comprising (i) claims about representing, reasoning with, and learning domain knowledge to support the construction of explanations; (ii) three fundamental axes to characterize explanations; and (iii) a methodology for constructing these explanations. Second, we describe an architecture for robots that implements this theory and supports scalability to complex domains and explanations. We demonstrate the architecture’s capabilities in the context of a simulated robot (a) moving target objects to desired locations or people; or (b) following recipes to bake biscuits.

View PDFchevron_right
Robots that change their world: Inferring Goals from Semantic Knowledge
Javier Gonzalez-Jimenez

A growing body of literature shows that endowing a mobile robot with semantic knowledge, and with the ability to reason from this knowledge, can greatly increase its capabilities. In this paper, we explore a novel use of semantic knowledge: we encode information about how things should be, or norms, to allow the robot to infer deviations from these norms and to generate goals to correct these deviations. For instance, if a robot has semantic knowledge that perishable items must be kept in a refrigerator, and it observes a bottle of milk on a table, this robot will generate the goal to bring that bottle into a refrigerator. Our approach provides a mobile robot with a limited form of goal autonomy: the ability to derive its own goals to pursue generic aims. We illustrate our approach in a full mobile robot system that integrates a semantic map, a knowledge representation and reasoning system, a task planner, as well as standard perception and navigation routines.

View PDFchevron_right
Belief modelling for situation awareness in human-robot interaction
Pierre Lison

2010

To interact naturally with humans, robots needs to be aware of their own surroundings. This awareness is usually encoded in some implicit or explicit representation of the situated context. In this research report, we present a new framework for constructing rich belief models of the robot's environment.

View PDFchevron_right

References (56)

  1. Aydemir, A., Göbelbecker, M., Pronobis, A., Sjöö, K., Jensfelt, P., Sep. 2011. Plan-based object search and exploration using semantic spatial knowledge in the real world. In: Proceedings of the European Conference on Mobile Robotics (ECMR'11). Örebro, Sweden, pp. 13-18.
  2. Aydemir, A., Pronobis, A., Gobelbecker, M., Jensfelt, P., 2013. Active visual object search in unknown environments using uncertain seman- tics. IEEE Transactions on Robotics 29 (4), 986-1002.
  3. Aydemir, A., Sjöö, K., Folkesson, J., Jensfelt, P., May 2011. Search in the real world: Active visual object search based on spatial relations. In: Proceedings of the 2011 IEEE International Conference on Robotics and Automation (ICRA 2011). Shanghai, China, pp. 2818-2824.
  4. Bäckström, C., Nebel, B., 1995. Complexity results for SAS + planning. Computational Intelligence 11 (4), 625-655.
  5. Brenner, M., Nebel, B., 2009. Continual planning and acting in dynamic multiagent environments. Autonomous Agents and Multi-Agent Systems 19, 297-331, 10.1007/s10458-009-9081-1.
  6. Christensen, H. I., Kruijff, G.-J. M., Wyatt, J. L. (Eds.), 2010. Cognitive Systems. Vol. 8 of Cognitive Systems Monographs. Springer, Berlin.
  7. Ekvall, S., Kragic, D., Jensfelt, P., 2007. Object detection and mapping for service robot tasks. Robotica 25 (2), 175-187.
  8. Galindo, C., Fernández-Madrigal, J.-A., González, J., Saffiotti, A., 2008. Robot task planning using semantic maps. Robotics and Autonomous Systems 56 (11), 955-966.
  9. Gardenfors, P., 1986. Belief revisions and the Ramsey test for conditionals. Philosophical Review 95, 81-93.
  10. Gerevini, A. E., Haslum, P., Long, D., Saetti, A., Dimopoulos, Y., 2009. Deterministic planning in the Fifth International Planning Competi- tion: PDDL3 and experimental evaluation of the planners. Artificial Intelligence 173 (5-6), 619-668.
  11. Göbelbecker, M., Gretton, C., Dearden, R., 2011. A switching planner for combined task and observation planning. In: Proceedings of the Twenty-Fifth AAAI Conference on Artificial Intelligence. San Francisco, USA, pp. 964-970.
  12. González-Banos, H., Latombe, J., 2001. A randomized art-gallery algorithm for sensor placement. In: SCG '01: Proceedings of the Seven- teenth Annual Symposium on Computational Geometry. ACM, New York, USA, pp. 232-240.
  13. Hanheide, M., Gretton, C., Dearden, R. W., Hawes, N. A., Wyatt, J. L., Pronobis, A., Aydemir, A., Göbelbecker, M., Zender, H., 2011. Exploiting probabilistic knowledge under uncertain sensing for efficient robot behaviours. In: Proceedings of the Twenty-Second International Joint Conference on Artificial Intelligence (IJCAI-11). Barcelona, Spain, pp. 2442-2449.
  14. Hanheide, M., Hawes, N., Wyatt, J. L., Göbelbecker, M., Brenner, M., Sjöö, K., Aydemir, A., Jensfelt, P., Zender, H., Kruijff, G.-J. M., 2010. A framework for goal generation and management. In: Proceedings of the AAAI Workshop on Goal-Directed Autonomy, WS-4 AAAI 2010. Atlanta, Georgia.
  15. Hawes, N., Hanheide, M., Hargreaves, J., Page, B., Zender, H., Jensfelt, P., 2011. Home alone: Autonomous extension and correction of spatial representations. In: Proceedings of the 2011 IEEE International Conference on Robotics and Automation (ICRA 2011). Shanghai, China, pp. 3907-3914.
  16. Hawes, N., Wyatt, J. L., 2010. Engineering intelligent information-processing systems with CAST. Advances in Engineering Informatics 24 (1), 27-39.
  17. Hawes, N., Wyatt, J. L., Sridharan, M., Jacobsson, H., Dearden, R., Sloman, A., Kruijff, G.-J., 2010. Architecture and representations. In: Cognitive Systems. Springer, Berlin, pp. 51-93.
  18. Helmert, M., 2006. The fast downward planning system. Journal of Artificial Intelligence Research 26, 191-246.
  19. Jacobsson, H., Hawes, N., Kruijff, G.-J., Wyatt, J., 2008. Crossmodal content binding in information-processing architectures. In: Proceedings of the 3rd ACM/IEEE International Conference on Human Robot Interaction -HRI '08. Amsterdam, Netherlands, pp. 81-88.
  20. Jain, D., Mosenlechner, L., Beetz, M., May 2009. Equipping robot control programs with first-order probabilistic reasoning capabilities. In: In Proceedings of 2009 IEEE International Conference on Robotics and Automation (ICRA 2009). IEEE, pp. 3626-3631.
  21. Janíček, M., 2012. Abductive reasoning for continual dialogue understanding. In: Slavkovik, M., Lassiter, D. (Eds.), New Directions in Logic, Language, and Computation. Springer, pp. 16-31.
  22. Kaplow, R., Atrash, A., Pineau, J., 2010. Variable resolution decomposition for robotic navigation under a pomdp framework. In: Proceedings of the 2010 IEEE International Conference on Robotics and Automation (ICRA 2010). Anchorage, USA, pp. 369-376.
  23. Klenk, M., Molineaux, M., Aha, D. W., 2013. Goal-driven autonomy for responding to unexpected events in strategy simulations. Computa- tional Intelligence 29 (2), 187-206.
  24. Kollar, T., Roy, N., 2009. Utilizing object-object and object-scene context when planning to find things. In: Proceedings of the 2009 IEEE International Conference on Robotics and Automation (ICRA'09). Kobe, Japan, pp. 4116-4121.
  25. Kraft, D., Bas ¸eski, E., Popović, M., Batog, A. M., Kjaer-Nielsen, A., Krüger, N., Petrick, R., Geib, C., Pugeault, N., Steedman, M., Asfour, T., Dillmann, R., Kalkan, S., Wörgötter, F., Hommel, B., Detry, R., Piater, J., 2008. Exploration and planning in a three-level cognitive architecture. In: International Conference on Cognitive Systems (CogSys 2008). Karlsruhe, Germany, pp. 71-78.
  26. Kunze, L., Beetz, M., Saito, M., Azuma, H., Okada, K., 2012. Searching objects in large-scale indoor environments: A decision-theoretic approach. In: Proceedings of the 2012 IEEE International Conference on Robotics and Automation (ICRA 2012). St. Paul, USA, pp. 4385- 4390.
  27. Lewis, D., 1973. Counterfactuals. Harvard University Press.
  28. Lienhart, R., Maydt, J., 2002. An extended set of Haar-like features for rapid object detection. In: Proceedings of the 2002 International Conference on Image Processing (ICIP 2002). Vol. 1. Rochester, New York, pp. 900-903.
  29. Martinez-Cantin, R., de Freitas, N., Brochu, E., Castellanos, J., Doucet, A., 2009. A Bayesian exploration-exploitation approach for optimal online sensing and planning with a visually guided mobile robot. Autonomous Robots 27 (2), 93-103.
  30. Mitchell, T. M., Keller, R. M., Kedar-Cabelli, S. T., 1986. Explanation-based generalization: A unifying view. Machine Learning 1 (1), 47-80.
  31. Mooij, J. M., Aug. 2010. libDAI: A free and open source C++ library for discrete approximate inference in graphical models. Journal of Machine Learning Research 11, 2169-2173.
  32. Peter Bonasso, R., James Firby, R., Gat, E., Kortenkamp, D., Miller, D. P., Slack, M. G., Apr. 1997. Experiences with an architecture for intelligent, reactive agents. Journal of Experimental & Theoretical Artificial Intelligence 9 (2-3), 237-256.
  33. Petrick, R. P. A., Bacchus, F., 2002. A knowledge-based approach to planning with incomplete information and sensing. In: Proceedings of the Sixth International Conference on Artificial Intelligence Planning Systems 2002. Toulouse, France, pp. 212-222.
  34. Pineau, J., Montemerlo, M., Pollack, M., Roy, N., Thrun, S., 2003. Towards robotic assistants in nursing homes: Challenges and results. Robotics and Autonomous Systems 42 (3), 271-281.
  35. Pronobis, A., Jensfelt, P., 2012. Large-scale semantic mapping and reasoning with heterogeneous modalities. In: Proceedings of the 2012 IEEE International Conference on Robotics and Automation (ICRA'12). St. Paul, USA, pp. 3515-3522.
  36. Pronobis, A., Mozos, O. M., Caputo, B., Jensfelt, P., February 2010. Multi-modal semantic place classification. Int. J. Robot. Res. 29 (2-3), 298-320.
  37. Pronobis, A., Sjöö, K., Aydemir, A., Bishop, A. N., Jensfelt, P., 2009. A framework for robust cognitive spatial mapping. In: Proceedings of the 14th International Conference on Advanced Robotics (ICAR'09). Munich, Germany, pp. 1-8.
  38. Pronobis, A., Sjöö, K., Aydemir, A., Bishop, A. N., Jensfelt, P., 2010. Representing spatial knowledge in mobile cognitive systems. In: 11th International Conference on Intelligent Autonomous Systems (IAS-11). Ottawa, Canada, pp. 133-142.
  39. Reiter, R., 1987. A theory of diagnosis from first principles. Artificial Intelligence 32 (1), 57-95.
  40. Richtsfeld, A., Mörwald, T., Zillich, M., Vincze, M., 2010. Taking in shape: Detection and tracking of basic 3D shapes in a robotics context. In: Proceedings of the 15th Computer Vision Winter Workshop. Nové Hrady, Czech Republic, pp. 91-98.
  41. Shanahan, M., 1989. Prediction is deduction but explanation is abduction. In: Proceedings of the 11th International Joint Conference on Artificial Intelligence (IJCAI'89). Vol. 2. Detroit, USA, pp. 1055-1060.
  42. Shanahan, M., 2005. Perception as abduction: Turning sensor data into meaningful representation. Cognitive Science 29 (1), 103-134.
  43. Simmons, R., Goodwin, R., Haigh, K. Z., Koenig, S., O'Sullivan, J., 1997. A layered architecture for office delivery robots. In: Proceedings of the First International Conference on Autonomous Agents (AGENTS '97). Marina del Rey, USA, pp. 245-252.
  44. Sjöö, K., Aydemir, A., Schlyter, D., Jensfelt, P., July 2010. Topological spatial relations for active visual search. Tech. Rep. TRITA-CSC-CV 2010:2 CVAP317, Centre for Autonomous Systems, KTH, Stockholm.
  45. Sjöö, K., Zender, H., Jensfelt, P., Kruijff, G.-J. M., Pronobis, A., Hawes, N., Brenner, M., 2010. The Explorer system. In: Christensen, H. I., Kruijff, G.-J. M., Wyatt, J. L. (Eds.), Cognitive Systems. Springer, Berlin, pp. 395-421.
  46. Skočaj, D., Kristan, M., Vrečko, A., Mahnič, M., Janíček, M., Kruijff, G.-J. M., Hanheide, M., Hawes, N., Keller, T., Zillich, M., Zhou, K., 2011. A system for interactive learning in dialogue with a tutor. In: Proceedings of the 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2011). San Francisco, USA, pp. 3387-3394.
  47. Sohrabi, S., Baier, J. A., McIlraith, S. A., 2011. Preferred explanations: Theory and generation via planning. In: Proceedings of the 25th AAAI Conference on Artificial Intelligence. San Francisco, USA, pp. 261-267.
  48. Suh, I. H., Lim, G. H., Hwang, W., Suh, H., Choi, J.-H., Park, Y.-T., 2007. Ontology-based multi-layered robot knowledge framework (OMRKF) for robot intelligence. In: Proceedings of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2007). San Diego, USA, pp. 429-436.
  49. Talamadupula, K., Benton, J., Kambhampati, S., Schermerhorn, P., Scheutz, M., December 2010. Planning for human-robot teaming in open worlds. ACM Transactions on Intelligent Systems and Technology 1, 14:1-14:24.
  50. Tenorth, M., Beetz, M., 2009. KNOWROB -knowledge processing for autonomous personal robots. In: Proceedings of the 2009 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2009). St. Louis, USA, pp. 4261-4266.
  51. Velez, J. J., Huang, A. S., Hemann, G. A., Roy, N., Posner, I., 2012. Modelling observation correlations for active exploration and robust object detection. Journal of Artificial Intelligence Research 44, 423-453.
  52. Wyatt, J. L., Aydemir, A., Brenner, M., Hanheide, M., Hawes, N., Jensfelt, P., Kristan, M., Kruijff, G.-J. M., Lison, P., Pronobis, A., Sjöö, K., Skočaj, D., Vrečko, A., Zender, H., Zillich, M., December 2010. Self-understanding and self-extension: A systems and representational approach. IEEE Transactions on Autonomous Mental Development 2 (4), 282 -303.
  53. Younes, H. L. S., Littman, M., 2004. PPDDL 1.0: An extension to PDDL for expressing planning domains with probabilistic effects. Tech. Rep. CMU-CS-04-167, School of Computer Science, Carnegie Mellon University, Pittsburgh, Pennsylvania.
  54. Zender, H., Kruijff, G.-J. M., Kruijff-Korbayová, I., 2009. Situated resolution and generation of spatial referring expressions for robotic assistants. In: Proceedings of the Twenty-First International Joint Conference on Artificial Intelligence (IJCAI-09). Pasadena, USA, pp. 1604-1609.
  55. Zender, H., Mozos, O. M., Jensfelt, P., Kruijff, G.-J. M., Burgard, W., 2008. Conceptual spatial representations for indoor mobile robots. Robotics and Autonomous Systems 56 (6), 493-502.
  56. Zhang, S., Sridharan, M., Wyatt, J., 2015. Integrating probabilistic graphical models and non-monotonic logical inference for robots. IEEE Transactions on Robotics 31 (3), 699-713.

Related papers

Towards an Architecture for Knowledge Representation and Reasoning in Robotics
Jeremy L Wyatt

Lecture Notes in Computer Science, 2014

This paper describes an architecture that combines the complemen- tary strengths of probabilistic graphical models and declarative programming to enable robots to represent and reason with qualitative and quantitative descrip- tions of uncertainty and domain knowledge. An action language is used for the architecture’s low-level (LL) and high-level (HL) system descriptions, and the HL definition of recorded history is expanded to allow prioritized defaults. For any given objective, tentative plans created in the HL using commonsense reasoning are implemented in the LL using probabilistic algorithms, and the correspond- ing observations are added to the HL history. Tight coupling between the levels helps automate the selection of relevant variables and the generation of policies in the LL for each HL action, and supports reasoning with violation of defaults, noisy observations and unreliable actions in complex domains. The architecture is evaluated in simulation and on robots moving objects in indoor domains.

View PDFchevron_right
What Happened and Why ? A Mixed Architecture for Planning and Explanation Generation in Robotics
Mohan Sridharan

2015

This paper describes a mixed architecture that couples the non-monotonic logical reasoning capabilities of a declarative language with probabilistic belief revision, enabling robots to represent and reason with qualitative and quantitative descriptions of knowledge and uncertainty. Incomplete domain knowledge, including information that holds in all but a few exceptional situations, is represented as a Answer Set Prolog (ASP) program. The answer set obtained by solving this program is used for inference, planning, and for jointly explaining (a) unexpected action outcomes; and (b) partial scene descriptions extracted from sensor input. For any given task, each action in the plan contained in the answer set is executed probabilistically. For each such action, observations extracted from sensor inputs perform incremental Bayesian updates to a probabilistic (belief) distribution over a relevant subset of the domain, committing high probability beliefs as statements to the ASP program. T...

View PDFchevron_right
KR $^ 3$: An Architecture for Knowledge Representation and Reasoning in Robotics
Mohan Sridharan

This paper describes an architecture that combines the complementary strengths of declarative programming and probabilistic graphical models to enable robots to represent, reason with, and learn from, qualitative and quantitative descriptions of uncertainty and knowledge. An action language is used for the low-level (LL) and high-level (HL) system descriptions in the architecture, and the definition of recorded histories in the HL is expanded to allow prioritized defaults. For any given goal, tentative plans created in the HL using default knowledge and commonsense reasoning are implemented in the LL using probabilistic algorithms, with the corresponding observations used to update the HL history. Tight coupling between the two levels enables automatic selection of relevant variables and generation of suitable action policies in the LL for each HL action, and supports reasoning with violation of defaults, noisy observations and unreliable actions in large and complex domains. The architecture is evaluated in simulation and on physical robots transporting objects in indoor domains; the benefit on robots is a reduction in task execution time of 39% compared with a purely probabilistic, but still hierarchical, approach.

View PDFchevron_right
REBA-KRL: Refinement-Based Architecture for Knowledge Representation, Explainable Reasoning and Interactive Learning in Robotics
Mohan Sridharan

European Conference on Artificial Intelligence, 2020

This paper provides an overview of an architecture based on the principle of step-wise refinement for knowledge representation, explainable reasoning, and interactive learning, in robotics. We briefly describe the individual components of the architecture, and emphasize how the interplay between representation, reasoning, and learning, can be exploited to enable reliable and efficient decision making in dynamic domains. I. MOTIVATION Robots collaborating with humans in complex domains have to reason with different descriptions of incomplete domain knowledge and uncertainty. These descriptions include commonsense knowledge, e.g., default statements such as "books are usually in the library" and "cereal boxes are typically in the kitchen", which hold true in all but a few exceptional circumstances, e.g., cookbooks are usually in the kitchen. At the same time, information extracted by processing noisy inputs from sensors is often associated with quantitative measures of uncertainty, e.g., statements such as "I am 90% certain I saw the robotics book in the office". In addition, any robot operating in dynamic domains will have to augment or revise its existing knowledge over time. Furthermore, for effective collaboration with humans, robots should be able to explain their decisions, the underlying knowledge and beliefs, and the experiences that informed these beliefs, in a suitable format and at a desired level of abstraction. This paper describes an architecture that seeks to support the capabilities described above by exploiting the complementary strengths of declarative logic programming, probabilistic reasoning, and interactive learning. Here, we briefly describe the architecture's components. We use the following illustrative domain to present some execution traces demonstrating the working of the architecture. Example. [Robot Assistant (RA) Domain] Consider a robot delivering target objects to particular people or places (library, off ice, workshop, kitchen) in an indoor domain. Each place may have instances of objects such as book and cup. Each human has a role (e.g., engineer, manager, sales). Objects are characterized by the attributes such as size and color. The following are some other details of interest:

View PDFchevron_right
A framework for endowing an interactive robot with reasoning capabilities about perspective-taking and belief management
Grégoire Milliez, Rachid Alami

In daily human interactions, spatial reasoning occupies an important place. In this paper we present a situation assessment reasoner that generates relevant symbolic information from the geometry of the environment with respect to relations between objects and human capabilities. The role of SPARK (SPAtial Reasoning and Knowledge) component is to permanently maintain a state of the world in order to provide a basis for the robot to plan, to act, to react and to interact. More precisely, we describe here the way the system manages the hypotheses to be able to handle such knowledge in a flexible manner. Equipped with such capabilities, a robot that will interact with humans should be able to extract, compute or infer these relations and capabilities in order to communicate and interact efficiently in a natural way. To illustrate our work, we will explain how the robot is able to manage and update agents beliefs and pass Sally-Anne test. This work is part of a broader effort to develop a complete decisional framework for human-robot interactive task achievement.

View PDFchevron_right

Related topics

  • Robotics
  • Artificial Intelligence
  • Robotics (Computer Science)
  • Decision Making
  • Mobile Robotics
  • Probabilistic Networks
  • Autonomous Robotics
  • Deductive reasoning
  • Reasoning about Uncertainty
  • Decision And Game Theory
  • Cognitive Robotics
  • Human-Robot Interaction
  • Decision Making Under Uncertainty
  • Automated reasoning
  • Decision Theory
  • Proof and Reasoning
  • Bayesian Networks
  • Bayesian estimation
  • Probabilistic Graphical Models
  • Applied Bayesian Statistics
  • Automated Planning
  • Inference to the Best Explanation
  • Bayesian Network
  • Knowledge Representation and Rea...
  • Bayesian Belief Networks
  • Automated Planning and Scheduling
    • 580 California St., Suite 400
    San Francisco, CA, 94104
    © 2025 Academia. All rights reserved