Bruce Bon

Real-time collision avoidance for position-controlled manipulators

A new approach to real-time collision avoidance for position-controlled conventional 6-DOF and dexterous 7-DOF arms is developed, and supportive experimental results are presented. The collision avoidance problem is formulated and solved as a position-based force control problem. Virtual forces representing the intrusion of the arm into the obstacle safety zone are computed in real time using a spring-damper model. These forces are then nullified by employing an outer force control loop which perturbs the Cartesian commands for the arm position control system. The approach is implemented and tested on a 7-DOF RRC arm and a set of experiments are conducted in the laboratory. These experiments demonstrate perturbations of the end-effector position and orientation, as well as the arm posture, in order to avoid impending collisions. The proposed approach is simple, computationally fast, requires minimal modification to the arm control system, and applies to whole-arm collision avoidance for 7-DOF arms.

Sensing and perception research for space telerobotics at JPL

A useful space telerobot for on-orbit assembly, maintenance, and repair tasks must have a sensing and perception subsystem which can provide the locations, orientations, and velocities of all relevant objects in the work environment. This function must be accomplished with sufficient speed and accuracy to permit effective grappling and manipulation. Appropriate symbolic names must be attached to each object for use by higher-level planning algorithms. Sensor data and inferences must be presented to the remote human operator in a way that is both comprehensible in ensuring safe autonomous operation and useful for direct teleoperation. Research at JPL toward these objectives is described.

On-line collision avoidance for the Ranger telerobotic flight experiment

This paper describes an online approach to whole-arm collision avoidance for the NASA Ranger Telerobotic Flight Experiment. Using a geometric world model, minimum distances and nearest points between parts of the active dexterous 7-DOF manipulator and potential obstacles are computed by the Ranger obstacle detection software. Obstacle data is then used to compute virtual repulsive forces which perturb the operator-commanded manipulator Cartesian motions in order to avoid collisions. The virtual forces are computed at the tool-tip for end-effector position perturbation, at the wrist for end-effector orientation perturbation, and on the upper and lower arm links for arm angle perturbation. The paper also describes the software development environment, utilizing a 3D graphical simulation and providing a graphical user interface for display and operator command. Results of several test runs illustrating end-effector position and orientation perturbation, as well as arm angle perturbation, are presented.

Interactive command building and sequencing for supervised autonomy

An operator interface for interactive command building and sequencing for supervised autonomy is described. The system is the ground operator control station of a local-remote telerobotics system being developed for time-delayed remote-control of Space Station Freedom robots. The system provides stereo graphics overlay on video with interactive update of the environment. The operator selects objects in the environment with which to interact, and skills to specify the task to be performed, such as grasping a module, or opening a door. The information needed by the skill to operate on a specific object is stored in a knowledge base. In more complex cases, such as inserting a grasped module into a receptacle or using a tool on an object, the parameterization is generated by the skill querying the knowledge base for both the tool being held and the object on which the operation is performed.<<ETX>>

Real-time model-based obstacle detection for the NASA Ranger Telerobot

This paper describes the approach and algorithms developed for real-time model-based obstacle detection and distance computation for the NASA Ranger Telerobotic Flight Experiment. Objects of interest, such as manipulator arms or the ranger vehicle and solar arrays, are modeled using a small set of component types: edges, polygonal faces and cylindrical links. Link positions are computed using standard forward kinematics, and distances between object components are computed directly using equations derived from geometry. Prioritized lists of potential obstacles for each manipulator link eliminate needless distance computations and assure that the most likely obstacles are checked, even if the computation is terminated early due to real-time constraints. A test program, utilizing a 3D graphical simulation and providing a graphical user interface for operator control, has been developed and used to test and demonstrate obstacle detection. An earlier paper (1996) described how the obstacle detection results are utilized for collision avoidance.

Real-Time Model-Based Vision System For Object Acquisition And Tracking

A useful space telerobot for on-orbit assembly, maintenance, and repair tasks must have a sensing and perception subsystem which can provide the locations, orientations, and velocities of all relevant objects in the work environment. This function must be accomplished with sufficient speed and accuracy to permit effective grappling and manipulation. Appropriate symbolic names must be attached to each object for use by higher-level planning algorithms. Sensor data and inferences must be presented to the remote human operator in a way that is both comprehensible in ensuring safe autonomous operation and useful for direct teleoperation. Research at JPL toward these objectives is described.

Mars Volatiles and Climate Surveyor Robotic Arm

The primary purpose of the Mars Volatiles And Climate Surveyor (MVACS) Robotic Arm is to support the other MVACS science instruments by digging trenches in the Martian soil; acquiring and dumping soil samples into the Thermal Evolved Gas Analyzer; positioning the Soil Temperature Probe in the soil; positioning the Robotic Arm Air Temperature Sensor at various heights above the surface, and positioning the Robotic Arm Camera for taking images of the surface, trench, soil samples, magnetic targets, and other objects of scientific interest within its workspace. In addition to data collected from the Robotic Arm sensors during science support operations, the Robotic Arm will perform experiments along with the other science instruments to yield additional information on Martian soil mechanics in the vicinity of the lander. The experiments include periodic imaging of dumped soil piles, surface scraping and soil chopping experiments, compaction tests, insertion of the various end-effector tools into the soil, and trench cave-in tests. Data from the soil mechanics experiments will yield information on Martian soil properties such as angle of repose, cohesion, bearing strength, and grain size distribution.

Multi-range traversability indices for terrain-based navigation

Presents novel measures of terrain traversability at three different ranges, namely local, regional and global traversability indices. The local traversability index is related, by a set of linguistic rules, to local obstacles and surface softness within a local perception range, measured by on-board sensors mounted on a robot. The rule-based regional traversability index is computed from the terrain roughness and slope that are expected from video images of the terrain within a regional perception range obtained by on-board cameras. The global traversability index is obtained from a terrain topographic map, and is based on natural or man-made surface features, such as mountains and craters, within the global perception range. Each traversability index is represented by four fuzzy sets with linguistic labels {poor, low, moderate, high}, corresponding to surfaces that are unsafe, moderately unsafe, moderately safe or safe for traversal, respectively. These indices are used to develop a behavior-based navigation strategy for a mobile robot traversing a challenging terrain. The traversability indices form a basis of three navigation behaviors, namely local, regional and global traversal behaviors. These behaviors are integrated with a goal-seeking behavior to ensure that the mobile robot reaches the goal safely while avoiding obstacles and impassable terrain segments. The paper concludes with an illustrative graphical simulation study.

Real-time collision avoidance for 7-DOF arms

The paper presents experimental results that demonstrate a new approach to real-time collision avoidance for 7-DOF arms. The collision avoidance problem is formulated and solved as a force control problem. Virtual forces opposing intrusion of the arm into the obstacle safety zone are computed in real time. These forces are then nullified by employing an outer feedback loop which perturbs the arm Cartesian commands for the inner position control system. The approach is implemented and tested on a 7-DOF RRC arm and a set of experiments are conducted in the laboratory. These experiments demonstrate perturbation of the end-effector position and orientation, as well as the arm configuration, in order to avoid impending collisions The approach is simple, computationally fast, requires minimal modification to the arm control system, and applies to whole-arm collision avoidance.

Experiments in real-time collision avoidance for dexterous 7-DOF arms

The paper presents experimental results that demonstrate a new approach to real-time collision avoidance for 7-DOF arms. The collision avoidance problem is formulated and solved as a force control problem. Virtual forces opposing intrusion of the arm into the obstacle safety zone are computed in real time. These forces are then nullified by employing an outer feedback loop which perturbs the arm Cartesian commands for the inner position control system. The approach is implemented and tested on a 7-DOF RRC arm and a set of experiments are conducted in the laboratory. These experiments demonstrate perturbations of the end-effector position and orientation, as well as the arm configuration, in order to avoid impending collisions. The approach is simple, computationally fast, requires minimal modification to the arm control system, and applies to whole-arm collision avoidance.