Shraga Shoval

Using coded signals to benefit from ultrasonic sensor crosstalk in mobile robot obstacle avoidance

One problem oftentimes observed with arrays of multiple ultrasonic sensors (sonars) in obstacle detection and avoidance systems is crosstalk. The paper presents a method in which data generated by crosstalk is actually used to generate more reliable and accurate object detection. This is accomplished by assigning a unique code to the signals emitted by each sonar, so that the source sonar can be identified even if its signals echo is received by another sonar. Using geometric interpolation between the various sonars increases the accuracy of the measurements, overcoming their limited resolution. Experimental results show the potential of our system for mobile robotics obstacle detection and avoidance, as well as for localization using map-matching techniques.

Micro-Simulation Model for Assessing the Risk of Vehicle–Pedestrian Road Accidents

Data on traffic accidents clearly point to road black spots, where the accident rate is always high. However, road safety research is still far from understanding why these particular places on a road are risky. The reason is the lack of sufficient knowledge on how pedestrians and drivers interact when facing a potentially dangerous traffic situation, and the lack of an integrated framework that relates the data on human behavior to real-world traffic situations. We attempt to tackle this problem by developing SAFEPED, a multi-agent microscopic three-dimensional (3D) simulation of vehicle and pedestrian dynamics at a black spot. SAFEPED is a test platform for evaluating experimentally estimated drivers’ and pedestrians’ behavioral rules, and estimating accident risks in different traffic situations. It aims to analyze the design of existing and future black spots and to assess alternative architectural and environmental solutions in order to identify maximally efficient safety countermeasures.

Implementation of a Kalman filter in positioning for autonomous vehicles, and its sensitivity to the process parameters

Modern autonomous vehicles are using more than one method for performing the positioning task. The most common positioning methods for indoor vehicles are odometry for relative positioning and triangulation for absolute positioning. In many cases a Kalman filter is required to merge the data from the positioning systems and determines the vehicle position based on error analysis of the measurements and calculation procedures. A Kalman filter is particularly advantageous for “on-the-fly” positioning, which is performed while the vehicle is in motion. This paper presents the implementation of a Kalman filter in “ROBI” — an AGV for material handling in a manufacturing environment. The performance of the filter in estimating the position of the AGV and the effect of motion parameters (speed, path curvature, beacon layout etc.) on filter accuracy are shown.

Landmark configuration for absolute positioning of autonomous vehicles

Landmark navigation is a common positioning method of material handling vehicles operating in industrial environments. The position is calculated based on relative measurements to landmarks, the locations of which are known in advance. This paper presents a nonlinear optimization model used to determine the position of the minimum number of beacons required by a shop floor to guarantee accurate and reliable performance of AGVs. This position is based on guidelines developed by a previous error analysis study. The model is later converted to a binary linear programming model which is used as a lower bound to a heuristic solution procedure. Finally, two examples are used to illustrate the procedure.

A Foothold Selection Algorithm for Spider Robot Locomotion in Planar Tunnel Environments

In this paper we present an algorithm, called the partitioned cubes gaiting (PCG) algorithm, for planning the foothold positions of spider-like robots in planar tunnels bounded by piecewise linear walls. The paper focuses on three-limb robots, but the algorithm generalizes to robots with a larger number of limbs. The input to the PCG algorithm is a geometric description of the tunnel, a lower bound on the amount of friction at the contacts, as well as start and target foothold positions. Using efficient convex programming techniques, the algorithm approximates the possible foothold positions as a collection of cubes in contact configuration space (c-space). Each cube represents a contact independent set of feasible three-limb postures. A graph structure induced by the cubes has the property that its edges represent feasible motion between neighboring sets of three-limb postures. This motion is realized by lifting one limb while the other two limbs brace the robot against the tunnel walls. A shortest-path search along the graph yields a three–two–three gait pattern that moves the robot from start to target using a minimum number of foothold exchanges. In practical environments the algorithm runs in time, which is linear in the number of tunnel walls and polynomial in the degree of cube approximation of contact c-space. Simulations as well as experiments demonstrate the PCG algorithm in tunnel environments.

Stability of a multi tracked robot traveling over steep slopes

Tracked vehicles can offer the best solution for operation over complex terrains, including difficult surfaces. Tracked configuration is often required to guarantee the best mobility for unrestricted, all-weather tactical operations. A safe and reliable motion requires the vehicle to maintain stable contact with the terrain while avoiding slippage. This becomes critical when traveling over complex surfaces, especially steep and slippery slopes. This paper deals with the problem of how to maintain a stable motion of a multi tracked vehicle traveling over steep slopes. Changes in the vehicles internal configuration adjust the center of gravity to guarantee a stable motion. The paper analyzes the acceptable positions for the center of gravity given the slopes inclination, and the friction between the vehicle and the surface. Geometric solution for a two dimensional model is first presented, followed by the linear programming solution of a 3D model. Experimental results, as well as a field test conducted on our multi-tracked autonomous vehicle verify the theoretical analysis and provide constraints for stable motion given the vehicle and terrain geometry and friction between the vehicles tracks and the surface.

Odometry and triangulation data fusion for mobile-Robots environment recognition

Abstract Autonomous vehicles usually use more than one positioning system to improve their position estimate. Some positioning systems are advantageous in certain types of environments, while others are more efficient in others. This paper describes a data fusion method, where the differences between measurements are used to identify the type of terrain through which the vehicle is traveling. In this system, position estimate by odometry is compared to that calculated by triangulation , and the differences are fed into a neural network. This neural network, which is pertained by a set of different terrain types, classifies the examined environment by matching it with the most similar environment it can “recognize”.

On the Passive Force Closure Set of Planar Grasps and Fixtures

In this paper we consider grasps and fixtures whose contacts react according to force—displacement laws consistent with friction constraints at the contacts. The passive force closure set of such grasps and fixtures is the set of external wrenches (forces and torques) that can act on the grasped object and be stably balanced by the contacts. An external wrench belongs to this set if it induces a feasible equilibrium whose basin of attraction contains the initial unperturbed grasp configuration. The paper focuses on planar grasps and fixtures having sharp-tipped fingers or fixels that satisfy linear force displacement laws. Using Morse theory, we characterize the number and stability type of k-contact equilibria induced by a given external wrench. Based on this analysis, we provide closed-form expressions for the passive force closure set of k-contact grasps and fixtures. Computation of the allowed external wrenches is illustrated with examples. Finally, the paper describes experiments verifying the passive force closure set on a spring-loaded grasping system.

Sensory redundant parallel mobile mechanism

This paper presents a novel design for a mobile robot based on the kinematics of parallel mechanisms. The robot consists of 3 legs, each equipped with an asynchronous driving unit. The legs are connected to the driving units with spherical joints and to the upper plate with a revolute joint. Three additional encoders, attached to the upper revolute joints provide redundant data. This data is used by a kinematic model for accurate estimation of the robots configuration and position in space, even in rough terrains, where conventional odometry fails. Simulation results show the advantages of the design, and suggest a method for detection of irregularities of surfaces in unknown environments.

Design of a Quadruped Robot for Motion with Quasistatic Force Constraints

This paper presents a novel design of a four-legged “spider” robot capable of moving in a wide range of two-dimensional tunnels. The robot moves in a quasistatic manner, by stably bracing itself against the tunnel walls while moving its free parts to the next position. The design has been strongly influenced by the recent immobilization theory of Rimon and Burdick (1998a, 1998b). The theory dictates the minimum number of limbs such a mechanism can have, as well as the curvature of the mechanism footpads. The class of tunnel geometries dictates other key parameters of the robot, such as limb dimensions and number of degrees of freedom of each limb. We review the relevant components of the immobilization theory and describe its implications for the robot design. Then we describe our choice of other key design parameters of the robot. The spider-like robot will move under a worst-case assumption of slippery tunnel walls, and we also describe a locomotion strategy for the robot under this assumption. Finally, we describe an immobilization-based control algorithm for executing the motion strategy. The robot has been built, and experiments verifying its robustness with respect to leg-placement errors are described.