Kapil Katyal

Followers

Following

Public Views

Interests

Uploads

Papers by Kapil Katyal

Learning a Group-Aware Policy for Robot Navigation

2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Human-aware robot navigation promises a range of applications in which mobile robots bring versat... more Human-aware robot navigation promises a range of applications in which mobile robots bring versatile assistance to people in common human environments. While prior research has mostly focused on modeling pedestrians as independent, intentional individuals, people move in groups; consequently, it is imperative for mobile robots to respect human groups when navigating around people. This paper explores learning group-aware navigation policies based on dynamic group formation using deep reinforcement learning. Through simulation experiments, we show that group-aware policies, compared to baseline policies that neglect human groups, achieve greater robot navigation performance (e.g., fewer collisions), minimize violation of social norms and discomfort, and reduce the robot's movement impact on pedestrians. Our results contribute to the development of social navigation and the integration of mobile robots into human environments.

Download

Meta Arcade: A Configurable Environment Suite for Meta-Learning

arXiv (Cornell University), Dec 1, 2021

Most approaches to deep reinforcement learning (DRL) attempt to solve a single task at a time. As... more Most approaches to deep reinforcement learning (DRL) attempt to solve a single task at a time. As a result, most existing research benchmarks consist of individual games or suites of games that have common interfaces but little overlap in their perceptual features, objectives, or reward structures. To facilitate research into knowledge transfer among trained agents (e.g. via multi-task and metalearning), more environment suites that provide configurable tasks with enough commonality to be studied collectively are needed. In this paper we present Meta Arcade, a tool to easily define and configure custom 2D arcade games that share common visuals, state spaces, action spaces, game components, and scoring mechanisms. Meta Arcade differs from prior environments in that both task commonality and configurability are prioritized: entire sets of games can be constructed from common elements, and these elements are adjustable through exposed parameters. We include a suite of 24 predefined games that collectively illustrate the possibilities of this framework and discuss how these games can be configured for research applications. We provide several experiments that illustrate how Meta Arcade could be used, including single-task benchmarks of predefined games, sample curriculum-based approaches that change game parameters over a set schedule, and an exploration of transfer learning between games.

Download

Complex Terrain Navigation via Model Error Prediction

2022 International Conference on Robotics and Automation (ICRA)

Robot navigation traditionally relies on building an explicit map that is used to plan collisionf... more Robot navigation traditionally relies on building an explicit map that is used to plan collisionfree trajectories to a desired target. In deformable, complex terrain, using geometric-based approaches can fail to find a path due to mischaracterizing deformable objects as rigid and impassable. Instead, we learn to predict an estimate of traversability of terrain regions and to prefer regions that are easier to navigate (e.g., short grass over small shrubs). Rather than predicting collisions, we instead regress on realized error compared to a canonical dynamics model. We train with an on-policy approach, resulting in successful navigation policies using as little as 50 minutes of training data split across simulation and real world. Our learningbased navigation system is a sample efficient shortterm planner that we demonstrate on a Clearpath Husky navigating through a variety of terrain including grassland and forest.

Download

High-Speed Robot Navigation using Predicted Occupancy Maps

2021 IEEE International Conference on Robotics and Automation (ICRA), 2021

Safe and high-speed navigation is a key enabling capability for real world deployment of robotic ... more Safe and high-speed navigation is a key enabling capability for real world deployment of robotic systems. A significant limitation of existing approaches is the computational bottleneck associated with explicit mapping and the limited field of view (FOV) of existing sensor technologies. In this paper, we study algorithmic approaches that allow the robot to predict spaces extending beyond the sensor horizon for robust planning at high speeds. We accomplish this using a generative neural network trained from real-world data without requiring human annotated labels. Further, we extend our existing control algorithms to support leveraging the predicted spaces to improve collision-free planning and navigation at high speeds. Our experiments are conducted on a physical robot based on the MIT race car using an RGBD sensor where were able to demonstrate improved performance at 4 m/s compared to a controller not operating on predicted regions of the map.

Download

Robust Policy Search for an Agile Ground Vehicle Under Perception Uncertainty

2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021

Learning robust policies for robotic systems operating in presence of uncertainty is a challengin... more Learning robust policies for robotic systems operating in presence of uncertainty is a challenging task. For safe navigation, in addition to the natural stochasticity of the environment and vehicle dynamics, the perception uncertainty associated with dynamic entities, e.g. pedestrians, must be accounted for during motion planning. To this end, we construct an algorithm with built-in robustness to uncertainty by directly minimizing an upper confidence bound on the expected cost of trajectories instead of employing a standard approach based on minimizing the expected cost itself. Perception uncertainty is incorporated into the policy search framework by predicting each pedestrian’s intent belief and propagating their state distribution in time using closed-loop goal-directed dynamics. We train the policy in simulation and show that it could be transferred to an agile ground vehicle for successful autonomous robot navigation in presence of pedestrians with perception uncertainty. We further show the superior performance of this policy over a policy that does not consider pedestrian intent and perception uncertainty.

Prediction-Based Uncertainty Estimation for Adaptive Crowd Navigation

Artificial Intelligence in HCI, 2020

Fast, collision-free motion through human environments remains a challenging problem for robotic ... more Fast, collision-free motion through human environments remains a challenging problem for robotic systems. In these situations, the robot's ability to reason about its future motion and other agents is often severely limited. By contrast, biological systems routinely make decisions by taking into consideration what might exist in the future based on prior experience. In this paper, we present an approach that provides robotic systems the ability to make future predictions of the environment. We evaluate several deep network architectures, including purely generative and adversarial models for map prediction. We further extend this approach to predict future pedestrian motion. We show that prediction plays a key role in enabling an adaptive, risk-sensitive control policy. Our algorithms are able to generate future maps with a structural similarity index metric up to 0.899 compared to the ground truth map. Further, our adaptive crowd navigation algorithm is able to reduce the number of collisions by 43% in the presence of novel pedestrian motion not seen during training.

Download

Learning to Imagine Manipulation Goals for Robot Task Planning

ArXiv, 2017

Prospection is an important part of how humans come up with new task plans, but has not been expl... more Prospection is an important part of how humans come up with new task plans, but has not been explored in depth in robotics. Predicting multiple task-level is a challenging problem that involves capturing both task semantics and continuous variability over the state of the world. Ideally, we would combine the ability of machine learning to leverage big data for learning the semantics of a task, while using techniques from task planning to reliably generalize to new environment. In this work, we propose a method for learning a model encoding just such a representation for task planning. We learn a neural net that encodes the k most likely outcomes from high level actions from a given world. Our approach creates comprehensible task plans that allow us to predict changes to the environment many time steps into the future. We demonstrate this approach via application to a stacking task in a cluttered environment, where the robot must select between different colored blocks while avoiding...

Download

The JHU/APL AI technology roadmap: research and exploratory development on the path to intelligent systems for every mission

Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications III, 2021

Brain-Machine Interface Development for Finger Movement Control

SpringerBriefs in Electrical and Computer Engineering, 2017

There have been many developments in brain-machine interfaces (BMI) for controlling upper limb mo... more There have been many developments in brain-machine interfaces (BMI) for controlling upper limb movements such as reaching and grasping. One way to expand the usefulness of BMIs in replacing motor functions for patients with spinal cord injuries and neuromuscular disorders would be to improve the dexterity of upper limb movements performed by including more control of individual finger movements. Many studies have been focusing on understanding the organization of movement control in the sensorimotor cortex of the human brain. Finding the specific mechanisms for neural control of different movements will help focus signal acquisition and processing so as to improve BMI control of complex actions. In a recently published study, we demonstrated, for the first time, online BMI control of individual finger movements using electrocorticography recordings from the hand area of sensorimotor cortex. This study expands the possibilities for combined control of arm movements and more dexterous hand and finger movements.

Face Detection and Object Recognition for a Retinal Prosthesis

We describe the recent development of assistive computer vision algorithms for use with the Argus... more We describe the recent development of assistive computer vision algorithms for use with the Argus II retinal prosthesis system. While users of the prosthetic system can learn and adapt to the limited stimulation resolution, there exists great potential for computer vision algorithms to augment the experience and significantly increase the utility of the system for the user. To this end, our recent work has focused on helping with two different challenges encountered by the visually impaired: face detection and object recognition. In this paper, we describe algorithm implementations in both of these areas that make use of the retinal prosthesis for visual feedback to the user, and discuss the unique challenges faced in this domain.

Group-Aware Robot Navigation in Crowded Environments

ArXiv, 2020

Human-aware robot navigation promises a range of applications in which mobile robots bring versat... more Human-aware robot navigation promises a range of applications in which mobile robots bring versatile assistance to people in common human environments. While prior research has mostly focused on modeling pedestrians as independent, intentional individuals, people move in groups; consequently, it is imperative for mobile robots to respect human groups when navigating around people. This paper explores learning group-aware navigation policies based on dynamic group formation using deep reinforcement learning. Through simulation experiments, we show that group-aware policies, compared to baseline policies that neglect human groups, achieve greater robot navigation performance (e.g., fewer collisions), minimize violation of social norms and discomfort, and reduce the robot’s movement impact on pedestrians. Our results contribute to the development of social navigation and the integration of mobile robots into human environments.

Download

Occupancy Map Prediction Using Generative and Fully Convolutional Networks for Vehicle Navigation

ArXiv, 2018

Fast, collision-free motion through unknown environments remains a challenging problem for roboti... more Fast, collision-free motion through unknown environments remains a challenging problem for robotic systems. In these situations, the robot's ability to reason about its future motion is often severely limited by sensor field of view (FOV). By contrast, biological systems routinely make decisions by taking into consideration what might exist beyond their FOV based on prior experience. In this paper, we present an approach for predicting occupancy map representations of sensor data for future robot motions using deep neural networks. We evaluate several deep network architectures, including purely generative and adversarial models. Testing on both simulated and real environments we demonstrated performance both qualitatively and quantitatively, with SSIM similarity measure up to 0.899. We showed that it is possible to make predictions about occupied space beyond the physical robot's FOV from simulated training data. In the future, this method will allow robots to navigate thro...

Download

The Modular Prosthetic Limb

Wearable Robotics, 2020

Abstract Revolutionizing Prosthetics is a government-sponsored program focused on maturing the ma... more Abstract Revolutionizing Prosthetics is a government-sponsored program focused on maturing the many foundational technologies that comprise neural prosthetic systems. Targeting the needs of amputees and movement-impaired individuals, the program focused on technological advancements in areas such as advanced neural recording devices, neural decoding and encoding algorithms, and upper limb prosthetic systems. A primary objective of the program was to support technological advancement of wearable prosthetic devices at the upper limb and hand level. The Modular Prosthetic Limb (MPL) is a representation of the vision for a highly anthropometric prosthetic limb. With 26 articulating joints, 17 actuators, and hundreds of internal sensors for feedback, the vision of the MPL was to replicate the dexterity, speed, and strength of the human hand to an extent never realized in a prosthetic device. Here we describe the MPL’s evolution over the past 13 years and describe the system in entirety, focusing on the fundamental characteristics of the system from hardware to software and controls. Additionally, we briefly touch upon some clinical applications and alternative use cases for the system to date.

Leveraging Deep Reinforcement Learning for Reaching Robotic Tasks

2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2017

This work leverages Deep Reinforcement Learning (DRL) to make robotic control immune to changes i... more This work leverages Deep Reinforcement Learning (DRL) to make robotic control immune to changes in the robot manipulator or the environment and to perform reaching, collision avoidance and grasping without explicit, prior and fine knowledge of the human arm structure and kinematics, without careful hand-eye calibration, solely based on visual/retinal input, and in ways that are robust to environmental changes. We learn a manipulation policy which we show takes the first steps towards generalizing to changes in the environment and can scale to new manipulators. Experiments are aimed at a) comparing different DCNN network architectures b) assessing the reward prediction for two manipulators with varying kinematics and c) performing a sensitivity analysis comparing a classical visual servoing formulation of the reaching task with the proposed DRL method.

Download

Visual Robot Task Planning

2019 International Conference on Robotics and Automation (ICRA), 2019

Prospection, the act of predicting the consequences of many possible futures, is intrinsic to hum... more Prospection, the act of predicting the consequences of many possible futures, is intrinsic to human planning and action, and may even be at the root of consciousness. Surprisingly, this idea has been explored comparatively little in robotics. In this work, we propose a neural network architecture and associated planning algorithm that (1) learns a representation of the world useful for generating prospective futures after the application of high-level actions, (2) uses this generative model to simulate the result of sequences of high-level actions in a variety of environments, and (3) uses this same representation to evaluate these actions and perform tree search to find a sequence of high-level actions in a new environment. Models are trained via imitation learning on a variety of domains, including navigation, pick-and-place, and a surgical robotics task. Our approach allows us to visualize intermediate motion goals and learn to plan complex activity from visual information.

Download

Nested marsupial robotic system for search and sampling in increasingly constrained environments

2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC), 2016

This paper presents a nested marsupial robotic system and its execution of a notional disaster re... more This paper presents a nested marsupial robotic system and its execution of a notional disaster response task. Human supervised autonomy is facilitated by tightly-coupled, high-level user feedback enabling command and control of a bimanual mobile manipulator carrying a quadrotor unmanned aerial vehicle that carries a miniature ground robot. Each robot performs a portion of a mock hazardous chemical spill investigation and sampling task within a shipping container. This work offers an example application for a heterogeneous team of robots that could directly support first responder activities using complementary capabilities of autonomous dexterous manipulation and mobility, autonomous planning and control, and teleoperation. The task was successfully executed during multiple live trials at the DARPA Robotics Challenge Technology Expo in June 2015. A key contribution of the work is the application of a unified algorithmic approach to autonomous planning, control, and estimation supporting vision-based manipulation and non-GPS-based ground and aerial mobility, thus reducing algorithmic complexity across this capability set. The unified algorithmic approach is described along with the robot capabilities, hardware implementations, and human interface, followed by discussion of live demonstration execution and results.

Intent-Aware Pedestrian Prediction for Adaptive Crowd Navigation

2020 IEEE International Conference on Robotics and Automation (ICRA), 2020

(JHU) to develop adaptive crowd navigation policies for robots by reasoning and predicting future... more (JHU) to develop adaptive crowd navigation policies for robots by reasoning and predicting future pedestrian motion. data sets on pedestrians and achieve comparable or better prediction accuracy compared with several stateof-the-art approaches (shown in Table 1). Moreover, we show that confidence in the prediction of pedestrian motion can be used to adjust the risk of a navigation policy adaptively to afford the most comfortable level as measured by the frequency of personal space violation in comparison with baselines. Furthermore, our adaptive navigation policy is able to reduce the number of collisions by 43% in the presence of novel pedestrian motion not seen during training. TECHNICAL APPROACH Machine learning has had a significant impact on many domains, including object recognition, natural language processing, and speech recognition. In recent years, we have seen a significant rise in the use of machine learning, and reinforcement learning specifically, for robotic navigation tasks. The advantage is that robotic systems are now capable of learning skills versus

Download

Uncertainty-Aware Occupancy Map Prediction Using Generative Networks for Robot Navigation

2019 International Conference on Robotics and Automation (ICRA), 2019

Efficient exploration through unknown environments remains a challenging problem for robotic syst... more Efficient exploration through unknown environments remains a challenging problem for robotic systems. In these situations, the robot’s ability to reason about its future motion is often severely limited by sensor field of view (FOV). By contrast, biological systems routinely make decisions by taking into consideration what might exist beyond their FOV based on prior experience. We present an approach for predicting occupancy map representations of sensor data for future robot motions using deep neural networks. We develop a custom loss function used to make accurate prediction while emphasizing physical boundaries. We further study extensions to our neural network architecture to account for uncertainty and ambiguity inherent in mapping and exploration. Finally, we demonstrate a combined map prediction and information-theoretic exploration strategy using the variance of the generated hypotheses as the heuristic for efficient exploration of unknown environments.

Autonomous Grasping Robotic Aerial System for Perching (AGRASP)

2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018

This paper presents an autonomous perching concept for multirotor aerial vehicles. The Autonomous... more This paper presents an autonomous perching concept for multirotor aerial vehicles. The Autonomous Grasping Robotic Aerial System for Perching (AGRASP)represents a novel integration of robotics perception, vision-based path planning, and biomimetically-inspired manipulation on a small, lightweight aerial robot with highly-constrained sensor and processing capacity. Computationally lightweight perception algorithms pull candidate perch structures out of a complex environment with no a priori knowledge of the operational space. The innovative manipulator design combines both active grasp and passive grip enabling it to maintain hold on the perch even with all power off. We experimentally demonstrate, for the first time, a quadrotor autonomously detecting and landing on a perch relying solely on onboard sensing and processing.

Eye Movement Control in the Argus II Retinal-Prosthesis Enables Reduced Head Movement and Better Localization Precision

Investigative ophthalmology & visual science, Feb 1, 2018

Visual scanning by sighted individuals is done using eye and head movements. In contrast, scannin... more Visual scanning by sighted individuals is done using eye and head movements. In contrast, scanning using the Argus II is solely done by head movement, since eye movements can introduce localization errors. Here, we tested if a scanning mode utilizing eye movements increases visual stability and reduces head movements in Argus II users. Eye positions were measured in real-time and were used to shift the region of interest (ROI) that is sent to the implant within the wide field of view (FOV) of the scene camera. Participants were able to use combined eye-head scanning: shifting the camera by moving their head and shifting the ROI within the FOV by eye movement. Eight blind individuals implanted with the Argus II retinal prosthesis participated in the study. A white target appeared on a touchscreen monitor and the participants were instructed to report the location of the target by touching the monitor. We compared the spread of the responses, the time to complete the task, and the amo...

Download

Kapil Katyal

Uploads

Papers by Kapil Katyal

Log In