For a complete list of my publications, click here.
For a complete list of my robot videos with descriptions, click here.

The major topics of my research are listed below, with the latest ones on top. Please click on the topic title for more details.

Integrated Planning and Control for Graceful Navigation of Balancing Mobile Robots

Traditionally, motion planning and control for mobile robots have been decoupled. Robot motion planning procedures, generally, account for obstacles in the environment and workspace constraints but do not account for the system dynamics and the constraints on them. They also do not have any knowledge of the details of the controller that is used to achieve these motion plans. On the other hand, the controller does not have any knowledge of the workspace constraints and obstacles in the environment. These decoupled approaches work well for kinematic systems. Though it is possible to make highly dynamic balancing mobile robots like the ballbot navigate environments using these decoupled procedures, they are often sub-optimal and result in jerky motions, where the controller is fighting with the dynamics of the system to move it around. Moreover, when disturbed, these procedures often either result in collision with the obstacles or drive the system unstable. In order to achieve robust, graceful and collision-free motions, an integrated planning and control procedure is necessary, where both the planner and the controller understand the system dynamics and also understand each other's details.

In this work, motions are planned in shape space that enable the balancing robots to achieve fast, graceful motions in position space, taking into account the dynamic constraint equations. Controllers called motion policies are designed to achieve fast, graceful motions in small domains of the position space that are collision-free. A hybrid control architecture is used for motion planning, where the planner chooses a sequence of motion policies to achieve the overall navigation task. The sequence of motion policies are chosen such that they are gracefully composable, i.e., they result in overall graceful motion when composed. This ensures that the high-level motion planner has knowledge of the low-level controller it uses. Each motion policy is designed such that when sequentially composed they result in overall graceful motion. This ensures that the low-level controllers understand what the high-level motion planner is trying to achieve thereby forming a truly integrated planning and control procedure.

Related Publications

Umashankar Nagarajan, George Kantor and Ralph Hollis, "Integrated Planning and Control for Graceful Navigation of Shape-Accelerated Underactuated Balancing Mobile Robots", Accepted for publication at the IEEE Conference on Robotics & Automation (ICRA), May 2012. PDF
Umashankar Nagarajan, George Kantor and Ralph Hollis, "Hybrid Control for Navigation of Shape-Accelerated Underactuated Balancing Systems", Proceedings of the IEEE Conference on Decision and Control (CDC), pp. 3566 - 3571, December 2010.
PDF IEEE Xplore

Shape Space Planning for Balancing Mobile Robots

Balancing mobile robots are underactuated systems with second-order, non-integrable constraints on their dynamics that restrict the family of feasible configuration trajectories. The configuration space of any dynamic system can be divided into position and shape space. The position variables represent the position of the system in the world, whereas the shape variables are those that affect the inertia matrix of the system and dominate the system dynamics. In balancing mobile robots, the strong coupling between the position and shape dynamics makes it impossible to ignore the shape dynamics while tracking desired motions in position space. Therefore, it is necessary to plan appropriate shape space motions to achieve the desired position space motions. This work focuses on shape-accelerated balancing mobile robots like the ballbot, wherein non-zero shape configurations result in accelerations in position space.

In this work, a shape trajectory planner was developed for shape-accelerated balancing mobile robots like the ballbot, which uses just the dynamic constraint equations to plan shape trajectories, which when tracked will result in approximate tracking of desired position trajectories. The planner exploits the relationship between shape changes and acceleration in position space to plan the appropriate shape trajectories. The planner can handle systems with high-dimensional shape space and can also handle cases where a subset of the shape variables is artificially constrained. In the case of ballbot, the shape planner can plan trajectories for the body and arm angles in order to achieve desired ball motions on the floor. The planner is significantly faster than the direct collocation methods in finding feasible trajectories that best approximate the desired position space motions. Moreover, since this approach uses only the dynamic constraint equations, a subset of the equations of the motion, it is more robust to modeling uncertainties in actuator mechanisms and nonlinear friction effects.

Related Publications

Umashankar Nagarajan, Byungjun Kim and Ralph Hollis, "Planning in High-dimensional Shape Space for a Single-wheeled Balancing Mobile Robot with Arms", Accepted for publication at the IEEE Conference on Robotics & Automation (ICRA), May 2012. PDF
Umashankar Nagarajan, "Dynamic Constraint-based Optimal Shape Trajectory Planner for Shape-Accelerated Underactuated Balancing Systems", Proceedings of Robotics: Science and Systems (RSS), June 2010. PDF RoboticsProceedings

The ballbot - Control, Planning and Physical Interaction

The ballbot is a human-sized robot that balances on a ball. It is an underactuated, omnidirectional balancing mobile robot, which can also rotate about its vertical axis (yaw motion). It uses a triad of legs to remain statically stable when powered off. The ball is actuated using a four-motor inverse mouse-ball drive.

This work dealt with the design of various controllers for dynamic balancing, stationkeeping, velocity control, yaw motion while balancing, and automatic transition between statically stable and dynamically stable states. This work also involved development of an offline trajectory planning algorithm that plans motion for the ballbot between static configurations, i.e., rest-rest motions. This work also explored some interesting human-robot physical interactions that can be achieved as a result of dynamic stability.

Related Publications

Umashankar Nagarajan, George Kantor and Ralph Hollis, "Trajectory Planning and Control of a Dynamically Stable Single Spherical Wheeled Mobile Robot", Proceedings of the IEEE International Conference on Robotics & Automation (ICRA), pp. 3743 - 3748, May 2009. PDF IEEE Xplore
Umashankar Nagarajan, Anish Mampetta, George Kantor and Ralph Hollis, "State Transition, Balancing, Station Keeping and Yaw Control for a Dynamically Stable Single Spherical Wheel Mobile Robot", Proceedings of the IEEE International Conference on Robotics & Automation (ICRA), pp. 998 - 1003, May 2009. PDF IEEE Xplore
Umashankar Nagarajan, George Kantor and Ralph Hollis, "Human-Robot Physical Interaction with Dynamically Stable Mobile Robots", Proceedings of the ACM/IEEE International Conference on Human-Robot Interaction (HRI), pp. 281 - 282, March 2009. PDF ACM

Safe Fall Control for Humanoid Robots

Humanoid robots are expected to share human environments in the future and it is important to ensure safety of their operation. The fall of a humanoid robot can be disastrous as it can seriously damage both the robot and the objects in its surroundings. If the fall is inevitable, then the humanoid robot must have a fall control strategy that will minimize damage to the surrounding objects and/or minimize damage to the robot. This work focuses on developing humanoid fall control strategies that change the direction of fall such that the falling robot avoids hitting the surrounding objects and falls in a safe region.

In this work, a novel planning and control framework was developed to control humanoid fall direction in the presence of multiple objects. The planner evaluates a list of control strategies and selects the best strategy that results in the "safest" estimated direction of fall. The control strategies including foot placement, whole body inertia shaping and partial body inertia shaping. The foot placement strategy changes the support base geometry, thereby changing the position and orientation of the leading tipping edge. The inertia shaping procedures use either all the joints or a subset of them to generate angular momentum in the desired fall direction. The fall planner is also able to select "No Action" as the best strategy, if appropriate. The fall performance is continuously tracked and can be improved in real-time. The effectiveness of this algorithm was demonstrated in simulation on an ASIMO-like humanoid robot for a variety of different cases.

Related Publications

Ambarish Goswami, Seung-kook Yun, Umashankar Nagarajan, Sung-Hee Lee, KangKang Yin and Shivaram Kalyanakrishnan, "Direction-Changing Fall Control in Humanoid Robots: Theory and Experiments", International Journal of Robotics Research (Under Review), 2011.
Umashankar Nagarajan and Ambarish Goswami, "Generalized Direction Changing Fall Control of Humanoid Robots among Multiple Objects", Proceedings of the IEEE International Conference on Robotics & Automation (ICRA), pp. 3316 - 3322, May 2010. PDF IEEE Xplore

Simulation Analysis and Control of Autonomous Bicycle

In this work, an autonomous bicycle system with a passive rider was modeled and simulated using MATLAB-SIMULINK. The system's inherent stability characteristics were studied using a number of simulation experiments. The system was subject to a variety of constraint trajectories for lean and steer and their contributions to the stability of the autonomous bicycle system were analyzed. It was found that a sinusoidal steer motion has a greater tendency to stabilize the bicycle at lower speeds than at higher speeds. Moreover, a higher frequency steer oscillation performed better than a lower frequency one. It was shown that the steer motion varies sinusoidally as the lean varies sinusoidally and the bicycle system has an inherent tendency to stabilize itself. This work also found that there is a stable velocity region where the bicycle system self stabilizes. It was originally claimed in literature that there is only a lower bound to this stable velocity region but we showed that there exists an upper bound too. This work also developed several fuzzy-logic and adaptive neuro-fuzzy controllers for stabilizing the autonomous bicycle system by controlling only the bicycle lean. The fuzzy rules were designed such that they exploit the inherent stability characteristics of the system and also the lean and steer relationship. These intelligent controllers were shown to successfully stabilize the autonomous bicycle system outside its stability velocity region.

Related Publications

Umashankar N and Himanshu Dutt Sharma, "Adaptive Neuro - Fuzzy Controller for Stabilizing Autonomous Bicycle", Proceedings of the IEEE International Conference on Robotics & Biomimetics (ROBIO), pp. 1652 - 1657, December 2006.
IEEE Xplore
Himanshu Dutt Sharma and Umashankar N, "A Robotic Model (ROBI) of Autonomous Bicycle System", International Conference on Computational Inteligence for Modelling Control and Automation (CIMCA), November 2006. IEEE Xplore
Himanshu Dutt Sharma and Umashankar N, "Fuzzy Controller Design for an Autonomous Bicycle System", Proceedings of IEEE International Conference on Engineering of Intelligent Systems (ICEIS), pp. 498 - 503, April 2006. IEEE Xplore
Himanshu Dutt Sharma and Umashankar N, "Stabilization of Autonomous Bicycle using Fuzzy Controller with Maximum Allowable Lean Constraint", Proceedings of the 3rd International Conference on Computational Intelligence, Robotics and Autonomous Systems (CIRAS), December 2005. PDF
Himanshu Dutt Sharma, Sameer M. Kale and Umashankar N. "Simulation Model for Studying Inherent Stability Characteristics of Autonomous Bicycle", Proceedings of IEEE International Conference on Mechatronics and Automation (ICMA), pp. 193 - 198, July 2005. IEEE Xplore

Umashankar Nagarajan