Research Projects

Resolution-Adaptive Risk-Aware Trajectory Planning for Unmanned Surface Vehicles

Operating in Congested Civilian Traffic

In this project, we developed a lattice-based, 5D trajectory planner for an unmanned surface vehicle (USV) operating in a marine environment with civilian boat traffic. The planner estimates the collision risk and reasons about the availability of contingency maneuvers to counteract unpredictable behaviors of civilian vessels. The planner also incorporates avoidance behaviors of civilian vessels into the search for a dynamically feasible trajectory to minimize collision risk. In order to be computationally efficient, the planner dynamically scales the control action primitives of a trajectory based on the distribution and concentration of civilian vessels, while preserving the dynamical feasibility of the primitives. We present a novel congestion metric to compare the complexity of different scenarios when evaluating the performance of the planner. Our results demonstrate that the risk-aware planner significantly decreases the number of collisions compared to a baseline, Velocity Obstacles (VO) based planner, especially in complex scenarios with a high number of civilian vessels. The adaptive capability of the planner improves its computational performance as compared to its non-adaptive version without sacrificing the quality of computed trajectories. This leads to a high replanning rate, which allows shorter traversal distances and smaller arrival times. For more details, see [10] in the list of journal publications.

Authors: Brual C. Shah, Petr Švec, Ivan R. Bertaska, Armando J. Sinisterra, Wilhelm Klinger, Karl von Ellenrieder, Manhar Dhanak, and Satyandra K. Gupta.

Adaptive Sampling Based, Dynamics-Aware Target Following for Unmanned Surface Vehicle

Operating under Coast Guard Collision Regulations in Civilian Traffic

In this project, we developed a model-predictive, local trajectory planning algorithm for an autonomous surface vehicle (USV) operating in congested and highly dynamic traffic. The planner generalizes the Velocity Obstacle concept to systems with non-linear dynamics, nonholonomic constraints, and any form of a low-level feedback controller. High planning performance is achieved by searching in a resolution-adaptive space of vehicle candidate motion goals for a motion goal that optimizes the arrival time to a given global goal. The algorithm gradually identifies, with increasing resolution and focused sampling, an approximate representation of spatio-temporal obstacle regions. The sampling of motion goals is constrained to ensure the International Regulations for Prevention of Collisions at Sea (COLREGs). We have carried out experiments to demonstrate the practical feasibility of the planner using a 14-foot (4.3 m) long catamaran USV operating in a harbor-like environment. For more details, see [24] in the list of conference publications.

Authors: Petr Švec, Brual C. Shah, Ivan R. Bertaska, Wilhelm Klinger, Armando J. Sinisterra, Karl von Ellenrieder, Manhar Dhanak, and Satyandra K. Gupta.

Dynamics-Aware Target Following for Unmanned Surface Vehicle

Operating under Coast Guard Collision Regulations in Civilian Traffic

In this project, we developed a model-predictive trajectory planning algorithm for following a target boat by an autonomous unmanned surface vehicle (USV; a self-driving boat) in an environment with static obstacle regions and civilian boats. The planner is capable of making a balanced trade-off among the following, possibly conflicting criteria: the risk of losing the target boat, trajectory length, risk of collision with obstacles, violation of the Coast Guard Collision Regulations (COLREGs), also known as "rules of the road", and execution of avoidance maneuvers against vessels that do not follow the rules. The planner addresses these criteria by combining a search for a dynamically feasible trajectory to a suitable pose behind the target boat in 4D state space, forming a time-extended lattice, and reactive planning that tracks this trajectory using control actions that respect the USV dynamics and are compliant with COLREGs. The reactive part of the planner represents a generalization of the Velocity Obstacles paradigm by computing obstacles in the control space using a system-identified, dynamic model of the USV as well as worst-case and probabilistic predictive motion models of other vessels. We present simulation and experimental results using an autonomous unmanned surface vehicle platform and a human-driven vessel to demonstrate that the planner is capable of fulfilling the above mentioned criteria. For more details, see [22 and 21] in the list of conference publications.

Authors: Petr Švec, Brual C. Shah, Ivan R. Bertaska, Jose Alvarez, Armando J. Sinisterra, Karl von Ellenrieder, Manhar Dhanak, and Satyandra K. Gupta.

Target Following with Motion Prediction for Unmanned Surface Vehicle

Operating in Cluttered Environments

The capability of following a moving target in an environment with obstacles is required as a basic and necessary function for realizing an autonomous unmanned surface vehicle (USV; a self-driving boat). Many target following scenarios involve follower and target vehicles with different physical capabilities. Moreover, the follower vehicle may not have prior information about the intended motion of the target boat. We have developed a novel approach for integrating motion goal prediction, trajectory planning, and trajectory tracking components to realize the capability of following through a close interplay of planning and control. Integrating these three components required adding new capabilities to each component to make them mutually compatible. The developed motion goal prediction algorithm enables computation of a suitable motion goal for the USV in the close vicinity to the target boat based on its motion, current state, as well as a distribution of obstacles. The trajectory planner generates a dynamically feasible, collision-free trajectory to allow the USV safely reach the desired motion goal. Since the USV is supposed to follow a moving target, the trajectory planning needs to be sufficiently fast and yet produce dynamically feasible and short trajectories. To target these challenges, the trajectory planner searches for trajectories through a hybrid, pose-position state space using a multi-resolution control action set. The underlying trajectory tracking controller computes desired surge speed for each segment of the trajectory and ensures that the vehicle maintains it. This video shows an autonomous USV following a teleoperated boat in a tank positioned inside the Neutral Buoyancy Research Facility (NBRF) at the University of Maryland. In this video, we do not demonstrate the motion goal prediction capability due to space constraints. A conference paper [17] and a journal version of the article [6] can be found in the list of publications.

Authors: Petr Švec, Atul Thakur, Eric Raboin, Brual C. Shah, and Satyandra K. Gupta.

Trajectory Planning for Unmanned Surface Vehicles to

Handle Sea Disturbances

In this project, we developed a trajectory planning algorithm that explicitly considers the probability of an unmanned surface vehicle (USV) deviating from its original course due to large ocean waves that can be only partially handled by a feedback controller. The computed trajectories are dynamically feasible and avoid static as well as dynamic obstacles. The video on the left shows a situation in which the vehicle considers it safe enough to move in between the two static obstacles under the sea state 1 (i.e., wave height is 0-0.1 m). On the contrary, the video on the right shows a situation in which the vehicle selects a safer but longer trajectory at the sea-state 4 (i.e., wave height is 1.25-2.5 m) and in the presence of an additional dynamic obstacle. The algorithm can adapt its search based on the user-specified risk thresholds and produces a contingency plan as a part of the computed trajectory. In case of USV mission, state transition probabilities need to be generated on-aboard. In order to make the computation fast, we have developed a model simplification algorithm based on temporal coherence and implemented it on GPU to accelerate its performance. For more details, see [3] in the list of journal publications and [11] in the list of conference publications.

Authors: Petr Švec, Atul Thakur, Max Schwartz, and Satyandra K, Gupta.

Automated Synthesis of Action Selection Policies for Unmanned Surface Vehicles

Operating in Adverse Environments

This video shows an autonomous unmanned surface vehicle (USV) executing maneuvers for blocking the advancement of an intruder boat toward a valuable target (oil tanker). This task requires the USV to utilize reactive planning complemented by short-term forward planning to generate specific maneuvers for blocking. The maneuvers are dictated by an action selection policy that was automatically synthesized and gradually refined by the developed evolutionary-based machine learning approach. The approach employs Genetic Programming (GP) for evolving sequences of commands as macro-actions, an iterative technique for automated detection of failure states of the vehicle, and generation of optimized, corrective sequences of action commands to preemptively avoid the failure states. The intruder is human-competitive and exhibits a deceptive behavior so that the USV cannot exploit regularity in its attacking behavior. We compared the performance of a hand-coded blocking policy to the performance of a policy that was automatically synthesized. Our results show that the performance of the automatically generated policy exceeds the performance of the hand-coded policy and thus demonstrates the feasibility of the proposed approach. For more details, see [2] in the list of journal publications and [8] in the list of conference publications.

Authors: Petr Švec and Satyandra K. Gupta. Visualization for the first portion of the video was made by Maxim Schwartz.

Model-Predictive Asset Defense by Team of Autonomous Surface Vehicles in Environment

with Civilian Boats

IMPORTANT: Please play the video in HD quality. This video illustrates a contract-based, decentralized planning approach for guarding a valuable asset by a team of autonomous unmanned surface vehicles (USV) against hostile boats in an environment with civilian traffic. The particular objective for the team of USVs is to maximize the expected time it takes a hostile boat to reach the asset. The team has to cooperatively deal with uncertainty about which boats poses an actual threat, employ active blocking to slow down the movement of boats towards the asset, and intelligently distribute themselves around the asset to optimize their future guarding opportunities. The developed planner incorporates a contract-based algorithm for allocating tasks to individual USVs through forward simulating the mission and assigning estimated utilities to candidate task allocation plans. The task allocation is based on marginal cost based contracting that allows decentralized, cooperative task negotiation among neighboring agents. The planner is capable of computing task allocation plans in real-time and is general enough to be used for a variety of scenarios. The underlying behaviors that correspond to individual tasks are optimized for two specific mission scenarios. We provide detailed analysis of simulation results and discuss the impact of communication interruptions, unreliable sensor data and simulation inaccuracies on performance. For more details, see [8] in the list of journal publications and [18] in the list of conference publications.

Authors: Eric Raboin, Petr Švec, Dana S. Nau, and Satyandra K. Gupta.

Turn Around Behavior Generation and Execution for Unmanned Ground Vehicles

Operating in Rough Terrain

Unmanned ground vehicles (UGVs) are emerging as an attractive alternative to manned vehicles in a wide variety of risky missions that require navigation over rugged terrains. In order to operate on rugged terrains, UGVs need to be able to rapidly turn around. This need arises because of limitations of perception accuracy and limited situational awareness. In many situations, executing turn around requires a large number of turning maneuvers in a tight space and it needs to be performed rapidly. So the goal of this work was to develop a turn around behavior and automatically generate its supporting data for the UGV to be able to guarantee real-time execution of a reliable N-point turn maneuver in a rugged terrain. This video demonstrates the developed approach for N-point turn planning and execution on an inclined platform. For more details, see [13] in the list of conference publications.

Authors: Madan M. Dabbeeru, Joshua D. Langsfeld, Petr Švec, and Satyandra K. Gupta.

Automated Indirect Manipulation of Irregular Shaped Cells with Optical Tweezers

Studying collective migration of cells is currently of considerable interest in biology and medicine leading to the possibility of novel diagnosis and treatments. Some cells are highly sensitive to direct laser exposure, which may influence their behavior or even cause photodamage. In addition, manual manipulation of cells is time-consuming making it hard to carry out systematic studies that are properly timed to exhibit the desired motility. This video demonstrates an automated indirect planning approach for precise, collision free, indirect manipulation of cells with irregular, dynamically changing shapes using Optical Tweezers (OT). The first portion of the video shows a collective migration of Dictyostelium cells on a glass surface. The later portion of the video shows a triplet pusher that is automatically manipulating an individual Dictyostelium cell to arrange it in a desired position and orientation. In the triplet pusher formation, two beads are directly trapped by laser and an intermediate bead (which is not directly trapped by laser) is used to ensure zero exposure of laser to the cell. For more details, see [7] in the list of journal publications and [13] in the list of conference publications.

Authors: Sagar Chowdhury, Atul Thakur, Petr Švec, Chenlu Wang, Wolfgang Losert, and Satyandra K. Gupta.

Automated Indirect Transport of Biological Cells using Planar Gripper Formations

Optical tweezers are used for manipulation of micron-sized dielectric beads and cells. Some biological cells are vulnerable to photo damage if subjected to laser-based direct manipulation. In such cases, precise manipulation of these cells can be accomplished by using gripper formations made up of glass beads, which are actuated by optical tweezers. Indirect manual manipulation of cells using grippers made of optically trapped beads is a time consuming process or sometimes just impossible. This video illustrates the developed approach for automated micromanipulation using gripper formations and heuristic based path planning for collision-free transport of biological cells. The objective of the heuristic planner is to transport the target cells to their respective goal positions in minimum time. Using the designed experiments, we evaluated the performance of different gripper formations in terms of gripper stability, speed of transport, and required laser power. For more details, see [5] in the list of journal publications and [15] in the list of conference publications.

Authors: Sagar Chowdhury, Atul Thakur, Petr Švec, Chenlu Wang, Wolfgang Losert, and Satyandra K. Gupta.

Automatic Manipulation of Biological Cells Inside Microfluidic Chamber Assisted by Optical Tweezers

This video presents an automated, physics-aware, planning approach for transporting biological cells in an optical tweezers-assisted microfluidic chamber. Fluid forces inside the chamber, modeled using computational fluid dynamics, are incorporated into the widely used Langevin equation to simulate the motion of the cells. We used the simulation to build a probability map of a cell successfully reaching one of the outlets of the chamber under the influence of fluid flow. The developed planner not only generates collision-free paths that exploit the fluid flow inside the chamber but also utilizes offline generated simulation data to decide suitable locations for releasing the cells. This ensures robust cell transport, while minimizing the required laser power and operational time. The planner is based on the heuristic D* Lite algorithm that employs a specific cost function for searching over a novel state-action space representation. For more details, see [4] in the list of journal publications and [9] in the list of conference publications.

Authors: Sagar Chowdhury, Petr Švec, Chenlu Wang, Kevin T. Seale, John P. Wikswo, Wolfgang Losert, and Satyandra K. Gupta.

In this video, we present a planning approach for automated high-speed transport of cells over large distances inside an Optical Tweezers (OT) assisted microfluidic chamber. The transport is performed in three steps that combine the optical trap and motorized stage motions. This includes optical trapping and transporting the cells to form a desired cell-ensemble that is suitable for a long distance transport, automatically moving the motorized stage to transport the cell-ensemble over a large distance while avoiding static obstacles, and distributing the cells from the ensemble to the desired locations using OT. The speeds of optical traps and the motorized stage are determined by modeling the motion of the particle under the influence of the optical trap. The desired cell-ensemble is automatically determined based on the geometry of the microfluidic chamber. We have developed a greedy heuristic method for optimal selection of the initial and the final location of the cell-ensemble to minimize the overall transport time while satisfying the constraints of the OT workspace. The approach is particularly useful in applications where cells are needed to be rapidly distributed inside a microfluidic chamber. For more details, see [20] in the list of conference publications.

Authors: Sagar Chowdhury, Atul Thakur, Chenlu Wang, Petr Švec, Wolfgang Losert, and Satyandra K. Gupta.

Automated Gripper Synthesis for Indirect Manipulation of Cells using Optical Tweezers

Optical Tweezers (OT) is used for highly accurate manipulations of biological cells. However, the direct exposure of cells to focused laser beam may negatively influence their biological functions. In order to overcome this problem, we generate multiple optical traps to grab and move a 3D ensemble of inert particles such as silica microspheres to act as a reconfigurable gripper for a manipulated cell. The relative positions of the microspheres are important in order for the gripper to be robust against external environmental forces and the exposure of high intensity laser on the cell to be minimized. In this paper, we present results of different gripper configurations, experimentally tested using our OT setup, that provide robust gripping as well as minimize laser intensity experienced by the cell. We developed a computational approach that allowed us to perform preliminary modeling and synthesis of the gripper configurations. The gripper synthesis is cast as a multi-objective optimization problem. For more details, see [12] in the list of conference publications.

Authors: Sagar Chowdhury, Petr Švec, Chenlu Wang, Wolfgang Losert, and Satyandra K. Gupta.