Christopher M. Clark

Altruistic Task Allocation despite Unbalanced Relationships within Multi-Robot Communities

Ryan Morton — Wed, 27 Jan 2010 09:31:33 PST

Typical Multi-Robot Systems consist of robots cooperating to maximize global fitness functions. However, in some scenarios, the set of interacting robots may not share common goals and thus the concept of a global fitness function becomes invalid. This work examines Multi-Robot Communities (MRC), in which individual robots have independent goals. Within the MRC context, we present a task allocation architecture that optimizes individual robot fitness functions over long time horizons using reciprocal altruism. Previous work has shown that reciprocating altruistic relationships can evolve between two willing robots, using market-based task auctions, while still protecting against selfish robots aiming to exploit altruism. As these relationships grow, robots are increasingly likely to perform tasks for one another without any reward or promise of payback. This work furthers this notion by considering cases where an imbalance exists in the altruistic relationship. The imbalance occurs when one robot can perform another robot's task, thereby exhibiting altruism, but the other robot cannot reciprocate since it is physically unable (e.g. lack of adequate sensors or actuators). A new altruistic controller to deal with such imbalances is presented. The controller permits a robot to build altruistic relationships with the community as a whole (one-to-many), instead of just with single robots (one-to-one). The controller is proven stable and guarantees altruistic relationships will grow, if robots are willing, while still minimizing the effects of selfish robots. Results indicate that the one-to-many controller performs comparable to the one-to-one on most problems, but excels in the case of an unbalanced altruistic relationship.

Towards Gaussian Multi-Robot SLAM for Underwater Robotics

Dave Kroetsch — Tue, 22 Sep 2009 11:23:09 PDT

This paper presents initial steps towards developing autonomous navigation capabilities for cooperating underwater robots. Specifically, Simultaneous Localization and Mapping, or SLAM, capabilities are investigated for a group of micro vehicles each equipped with a single downward facing camera and an Inertial Measurement Unit (IMU). To verify the approach, simulations of the multi-robot SLAM running in a 3D environment were conducted, where vehicles in close proximity of one another exchange maps to improve localization.

Autonomous Control for a Differential Thrust ROV

Wei Wang — Tue, 22 Sep 2009 11:23:08 PDT

Smaller autonomous underwater vehicles that use differential thrust for surge and yaw motion control has the advantage of low cost and, at the same time, increased maneuverability in yaw direction. However, since such vehicles are underactuated vehicles, design of an autonomous control system that enables the vehicle to autonomously track a predefined trajectory is challenging.In this paper, we presented such an autonomous control system and implemented it on a small underactuated ROV with the use of unscented Kalman filter for vehicle localization, a underwater acoustic positioning system as the position sensor and a compass as the direction sensor. In designing the control law, the integrator backstep technique is used to achieve Lyapunov stability. Computer simulation and field tests have shown that the autonomous control system works well for the vehicle to track a predefined trajectory and the the tracking error converged to a certain small value.

Robotic System Sensitivity to Neural Network Learning Rate: Theory, Simulation, and Experiments

Christopher M. Clark — Fri, 21 Aug 2009 09:28:00 PDT

Selection of neural network learning rates to obtain satisfactory performance from neural network controllers is a challenging problem. To assist in the selection of learning rates, this paper investigates robotic system sensitivity to neural network (NN) learning rate. The work reported here consists of experimental and simulation results. A neural network controller module, developed for the purpose of experimental evaluation of neural network controller performance of a CRS Robotics Corporation A460 robot, allows testing of NN controllers using real-time iterative learning. The A460 is equipped with a joint position proportional, integral, and derivative (PID) controller. The neural network module supplies a signal to compensate for remaining errors in the PID-controlled system. A robot simulation, which models this PID-controlled A460 robot and NN controller, was also developed to allow the calculation of sensitivity to the NN learning rate. This paper describes the implementation of three NN architectures: the error back-propagation (EBP) NN, mixture of experts (ME) NN, and manipulator operations using value encoding (MOVE) NN. The sensitivity of joint trajectory error of three NN controllers to learning rate was investigated using both simulation and experimentation. Similar results were obtained from the robot experiments and the dynamic simulation. These results of state sensitivity to NN learning rate confirm that the MOVE NN is least sensitive to learning rate, implying that selection of suitable learning rates for this NN architecture for the system considered is accomplished more readily than other NN architectures.

Dynamic Networks for Motion Planning in Multi-Robot Space Systems

Christopher M. Clark — Tue, 18 Aug 2009 12:42:55 PDT

A new motion planning framework is presented that enables multiple mobile robots with limited ranges of sensing and communication to maneuver and achieve goals safely in dynamic environments. The framework is applicable to both planetary rover and free-floating space robot applications. To combine the respective advantages of centralized and de-centralized planning, this framework is based on the concept of centralized planning within dynamic robot networks. As the robots move in their environment, localized robot groups form networks, within which world models and robot goals can be shared. Whenever a network is formed, new information becomes available to all robots in this network. With this new information, each robot uses a fast, centralized planner to compute new coordinated trajectories on the fly. Planning over multiple robot networks is decentralized and distributed. The applicability of the framework to planetary rovers is demonstrated in both simulations and real robot experiments. Also, the framework's applicability to free-floating robots in a 3D space environment is demonstrated in simulation.

Randomized Motion Planning for Groups of Nonholonomic Robots

Christopher M. Clark — Tue, 18 Aug 2009 12:42:54 PDT

This paper presents a technique for motion planning which is capable of planning trajectories for a large number of nonholonomic robots. The robots plan within a two dimensional environment that consists of stationary/moving obstacles, and fixed boundaries. Each robot uses randomized motion planner techniques based on Probabilistic Road Maps (PRM's) to construct it's own trajectory that is free of collisions with moving obstacles and other robots. The randomized motion planner allows easy integration of the robots nonholonomic constraint into the planning so that only kinematically consistent plans are constructed. It is important to include this constraint in the planning problem since the majority of planetary surface robots are nonholonomic. The speed of the road map construction allows planning in real-time, enabling the robot to maneuver safely in a dynamic environment. Communication between robots is infrequent since robots only communicate on a "need to know basis". To verify the planner's effectiveness, it was tested using both simulation and experiment.

Dynamic Robot Networks: A Coordination Platform for Multi-Robot Systems

Christopher M. Clark — Wed, 12 Aug 2009 17:08:18 PDT

A large number of tasks, from manufacturing to planetary exploration, have been successfully accomplished using single robot systems. Many of these tasks could be completed faster, more reliably, and on a larger scale using a cooperating team of autonomous mobile robots. However, robots must be able to coordinate their actions before cooperation is possible.This work aims to enable robots with the ability to coordinate their actions for safe navigation in dynamic, unknown environments. Specifically, the work focuses on: 1) the coordination of multiple robots when sensing and inter-robot communication are limited and 2) multi-robot motion planning in dynamic, unknown environments.First, a new coordination platform is introduced - Dynamic Robot Networks - that facilitates centralized robot coordination across ad hoc networks. As robots move about their environment, they dynamically form communication networks. Within these networks, robots can share local sensing information and coordinate the actions of all robots in the network.Second, a fast motion planner called within robot networks is presented. The planner is a probabilistic roadmap (PRM) motion planner augmented with new sampling strategies. These strategies decrease the planner's run time to enable on-the-fly planning - a key requirement for navigation in environments that are unknown a priori and contain moving obstacles.Simulations and real robot experiments are presented that demonstrate: 1) centralized robot coordination across dynamic robot networks, 2) on-the-fly motion planning to avoid moving and previously unknown obstacles, and 3) autonomous robot navigation towards individual goal locations.

Altruistic Relationships for Optimizing Task Fulfillment in Robot Communities

Christopher M. Clark — Wed, 12 Aug 2009 17:06:00 PDT

This paper introduces the concept of a multi-robot community in which multiple robots must fulfill their individual tasks while operating in a shared environment. Unlike typical multi-robot systems in which global cost functions are minimized while accomplishing a set of global tasks, the robots in this work have individual tasks to accomplish and individual cost functions to optimize (e.g. path length or number of objects to gather). A strategy is presented in which a robot may choose to aid in the completion of another robot's task. This type of "altruistic" action leads to evolving altruistic relationships between robots, and can ultimately result in a decrease in the individual cost functions of each robot. However, altruism with respect to another robot must be controlled such that it allows a relationship where both robots are altruistic, but protects an altruistic robot against a selfish robot that does not help others. A quantitative description of this altruism is presented, along with a law for controlling an individuals altruism. With a linear model of the altruism dynamics, altruistic relationships are proven to grow when robots are altruistic, but protect an altruistic robot from a selfish robot. Results of task planning simulations are presented that highlight the decrease in individual robot cost functions, as well as evolutionary trends of altruism between robots.

Probabilistic Road Map Sampling Strategies for Multi-Robot Motion Planning

Christopher M. Clark — Wed, 12 Aug 2009 17:05:59 PDT

This paper presents a Probabilistic Road Map (PRM) motion planning algorithm to be queried within Dynamic Robot Networks--a multi-robot coordination platform for robots operating with limited sensing and inter-robot communication.First, the Dynamic Robot Networks (DRN) coordination platform is introduced that facilitates centralized robot coordination across ad hoc networks, allowing safe navigation in dynamic, unknown environments. As robots move about their environment, they dynamically form communication networks. Within these networks, robots can share local sensing information and coordinate the actions of all robots in the network.Second, a fast single-query Probabilistic Road Map (PRM) to be called within the DRN platform is presented that has been augmented with new sampling strategies. Traditional PRM strategies have shown success in searching large configuration spaces. Considered here is their application to on-line, centralized, multiple mobile robot planning problems. New sampling strategies that exploit the kinematics of non-holonomic mobile robots have been developed and implemented. First, an appropriate method of selecting milestones in a PRM is identified to enable fast coverage of the configuration space. Second, a new method of generating PRM milestones is described that decreases the planning time over traditional methods. Finally, a new endgame region for multi-robot PRMs is presented that increases the likelihood of finding solutions given difficult goal configurations.Combining the DRN platform with these new sampling strategies, on-line centralized multi-robot planning is enabled. This allows robots to navigate safely in environments that are both dynamic and unknown. Simulations and real robot experiments are presented that demonstrate: (1) speed improvements accomplished by the sampling strategies, (2) centralized robot coordination across Dynamic Robot Networks, (3) on-the-fly motion planning to avoid moving and previously unknown obstacles and (4) autonomous robot navigation towards individual goal locations.

A Decentralized Reinforcement Learning Controller For Collaborative Driving

Luke Ng — Wed, 12 Aug 2009 17:05:58 PDT

Research in the collaborative driving domain strives to create control systems that coordinate the motion of multiple vehicles in order to navigate traffic both efficiently and safely. In this paper a novel individual vehicle controller based on reinforcement learning is introduced. This controller is capable of both lateral and longitudinal control while driving in a multi-vehicle platoon. The design and development of this controller is discussed in detail and simulation results showing learning progress and performance are presented.

Complete and Scalable Multi-Robot Planning in Tunnel Environments

Mike Peasgood — Wed, 12 Aug 2009 17:05:58 PDT

This paper addresses the challenging problem of finding collision-free trajectories for many robots moving to individual goals within a common environment. Most popular algorithms for multi-robot planning manage the complexity of the problem by planning trajectories for robots sequentially; such decoupled methods may fail to find a solution even if one exists. In contrast, this paper describes a multi-phase approach to the planning problem that guarantees a solution by creating and maintaining obstacle-free paths through the environment as required for each robot to reach its goal. Using a topological graph and spanning tree representation of a tunnel or corridor environment, the multi-phase planner is capable of planning trajectories for up to r = L-1 robots, where the spanning tree includes L leaves. Monte Carlo simulations in a large environment with varying number of robots demonstrate that the algorithm can solve planning problems requiring complex coordination of many robots that cannot be solved with a decoupled approach, and is scalable with complexity linear in the number of robots.

Motion Planning for Formations of Mobile Robots

T. D. Barfoot — Wed, 12 Aug 2009 17:05:57 PDT

This paper is concerned with planning the motion of mobile robots in formation, which means certain geometrical constraints are imposed on the relative positions and orientations of the robots throughout their travel. Specifically, a method of planning motion for formations of mobile robots with non-holonomic constraints is presented. The kinematic equations developed allow a certain class of formations to be maintained while the group as a whole exhibits motion. The work was validated using the Stanford Micro-Autonomous RoverS Testbed.

Development of a Microscopic Traffic Simulator for Inter-Vehicle Communication Application Research

Keith Yu Kit Leung — Mon, 10 Aug 2009 16:00:24 PDT

This paper describes the development of a microscopic traffic simulator purposely designed for ITS researchers studying inter-vehicle communication (IVC) concepts and applications in large traffic networks. The simulator can represent real life vehicles within the simulation by using data from vehicle Global Positioning System (GPS) receivers, enabling validation of theories with real vehicle data. The software is developed on top of the existing microscopic traffic simulator VISSIM with the added flexibility of modelling and efficiently handling communication between large numbers of vehicles. This along with the software architecture was discussed in detail.

A Genetic Algorithm Approach to Solve for Multiple Solutions of Inverse Kinematics Using Adaptive Niching and Clustering

Saleh Tabandeh — Mon, 10 Aug 2009 16:00:24 PDT

Inverse kinematics is a nonlinear problem that may have multiple solutions. A Genetic Algorithm(GA) for solving the inverse kinematics of a serial robotic manipulator is presented. The algorithm is capable of finding multiple solutions of the inverse kinematics through niching methods. Despite the fact that the number and position of solutions in the search space depends on the the position and orientation of the end-effector as well as the configuration of the robot, the number of GA parameters that must be set by a user are limited to a minimum through the use of an adaptive niching method. The only requirement of the algorithm is the forward kinematics equations which can be easily obtained from the link parameters and joint variables of the robot. For identifying and processing the outputs of this GA, a modified filtering and clustering phase is also added to the algorithm. The algorithm was tested to solve the inverse kinematic problem of a 3 degree-of-freedom(DOF) robotic manipulator.

Optimized Lane Assignment Using Inter-Vehicle Communication

Thanh-Son Dao — Mon, 10 Aug 2009 16:00:23 PDT

This paper presents an approach to lane assignment for highway vehicles that increases traffic throughput while ensuring they exit successfully at their destinations. Most of current traffic management systems do not consider lane organization of vehicles and only regulate traffic flows by controlling traffic signals or ramp meters. However, traffic throughput and efficient use of highways can be increased by coordinating driver behaviors intelligently. The goal of this research is to form a distributed control strategy for cars themselves to select lanes using inter-vehicle communication. Initial results are promising and demonstrate that intelligent lane selection can decrease vehicle traffic time.

Autonomous Fish Tracking by ROV Using Monocular Camera

Jun Zhou — Mon, 10 Aug 2009 16:00:22 PDT

This paper concerns the autonomous tracking of fish using a Remotely Operated Vehicle (ROV) equipped with a single camera. An efficient image processing algorithm is presented that enables pose estimation of a particular species of fish - a Large Mouth Bass. The algorithm uses a series of filters including the Gabor filter for texture, projection segmentation, and geometrical shape feature extraction to find the fishes distinctive dark lines that mark the body and tail. Feature based scaling then produces the position and orientation of the fish relative to the ROV. By implementing this algorithm on each frame of a series of video frames, successive relative state estimates can be obtained which are fused across time via a Kalman Filter. Video taken from a VideoRay MicroROV operating within Paradise Lake, Ontario, Canada was used to demonstrate off-line fish state estimation. In the future, this approach will be integrated within a closed-loop controller that allows the robot to autonomously follow the fish and monitor its behavior.

Localization of Multiple Robots with Simple Sensors

Mike Peasgood — Mon, 10 Aug 2009 16:00:21 PDT

This paper presents a distributed particle filter algorithm for localizing multiple mobile robots that are equipped only with low cost/low power sensors. This method is applicable to multi-micro robot systems, where size limitations restrict sensor selection (e.g. small infrared range finders). Localization of three robots in a known environment is conducted by combining measurements from a small number of simple range sensors with inter-robot distances obtained through an acoustic range finder system. The localization problem is formulated as estimating the global position and orientation of a single triangle, where corners of the triangle represent the positions of robots. The robot positions relative to the centroid of the triangle are then determined by trilateration using the inter-robot distance measurements. Each robot uses an identical particle filter algorithm to estimate the global position of the triangle. The best estimates determined by each particle filter are distributed among the robots for use in the following iteration. Simulations demonstrate the ability to perform global localization of three robots, each using a compass and two range finders. The results illustrate that this method can globally localize the robot team in a simulated indoor environment The results are compared to simulations where robots have access to only their own sensor data, which are unable to successfully localize under equivalent conditions.

Toward Systematic Approaches to Design and Implement Vehicles Semi-Active Control Systems

Hamidreza Bolandhemmat — Mon, 10 Aug 2009 16:00:20 PDT

In this paper a systematic while practical methodology has been presented for design of vehiclepsilas semiactive suspension systems. The semi-active control strategies developed to improve vehicle ride comfort and stability generally have a switching nature. This makes the design of the controlled suspension systems difficult and highly dependent on an extensive trial and error process. The proposed methodology maps the discontinuous control system model to a continuous linear region where all the time/frequency design techniques established in the conventional control system theory can be applied. If the semiactive control system is designed to satisfy some ride/stability requirements, an inverse mapping offers control law. The effectiveness of the proposed design methodology in dealing with real industrial problems is demonstrated with experimental results.

Applying Kinodynamic Randomized Motion Planning with a Dynamic Priority System to Multi-Robot Space Systems

Christopher M. Clark — Mon, 10 Aug 2009 16:00:19 PDT

Presents a new motion planning system that can construct collision-free trajectories for groups of robots in dynamic environments, without global knowledge or high-bandwidth communication. The robots plan within a confined environment that consists of stationary and moving obstacles. Each robot plans independently using Kinodynamic Randomized Motion Planner techniques to construct its own trajectory that is free of collisions with moving obstacles and other robots. To resolve conflicts between robot trajectories, a new Dynamic Priority System (DPS) is introduced which gives the right of way to the robot whose local workspace is the most crowded. The kinodynamic randomized motion planner allows easy integration of the robots nonholonomic constraint into the planning so that only kinematically and dynamically consistent plans are constructed. The speed of the trajectory construction allows planning in real-time, enabling the robot to maneuver safely in a dynamic environment. To verify the planner's effectiveness, it was tested using both simulation and experiment.

Dynamics of Step-Climbing with Deformable Wheels and Applications for Mobile Robotics

Alexander Wilhelm — Mon, 10 Aug 2009 16:00:18 PDT

Wheeled-mobile robots operating in human environments typically encounter small steps. Surmounting steps is normally not considered when determining peak torque needs, yet it can be the maximum requirement. This work looks at the statics and dynamics of this situation to determine the necessary peak torque. It finds that using a dynamic model that includes the wheel elasticity is essential for properly representing a real-world tire. When torque is increased using a step function, energy is stored in the tire-higher tire elasticity eases climbing. Knowledge of this phenomenon could facilitate the use of smaller actuators. The model is numerically integrated and results are found to agree with experiment.