In this paper, a collision avoidance method for robot manipulators is proposed. First, the potential function for avoiding obstacles is defined in the way that its value should be always null on the middle planes among obstacles. Next, the candidate graph of desired trajectory of end-effector is asked on such planes. We call it‘Safety First Graph’. The path on this graph is determined by the A algorithm. By the proposed method, the interference between the potential field for avoiding obstacles and that for approaching the end-effector to the destination position becomes very small. Therefore, the failure of search for a solution, due to the deadlock or the collision with obstacles, could be reduced in most cases. Moreover, re-searching the solution becomes possible based on the global desired trajectory. The method can be applied to general manipulators without paying attention to their kinematic features, and imposes no restriction on the motions. The effectiveness of the method has been verified by simulation studies.

View full abstract

Parallel manipulator has a unique structure whose six actuators constrain its end eff ector in parallel. A simple symmetrical design of the parallel manipulator leads to an octahedron, where a top triangle and a base triangle are connected by six length-controllable cylinders. Controlling the length of those cylinders, the top plate can be located at any position and orientation with respect to the base. This structure is fairly simple, but it has some practical problems concerning its working volume and the backdrivability of the mechanism. In order to solve those difficulties, our design employs three sets of pantograph link mechanism, each of which is driven by a couple of DC motors instead of linear actuators. It provides us a compact 6 DOF manipulator with high output/weight ratio. This report describes mechanical design, kinematic analysis, static force analysis and some mechanical charactoristics. In addition it discusses the singular pointsof the parallel manipulator.

View full abstract

A fast and general interference check algorithm which does not depend on the complexity of shape of an environment for a robot but on the distance between the robot and its nearest obstacle in the environment is proposed, which is based on the adoption of B-reps. as a model of the robot and octree as a model of the environment.
The octree is a natural hierarchical solid model that can change the resolution in positioning adaptively in reference to a region in the world space and is adequate to the environment model. Further, the B-reps. can represent easily a complex motion including rotation of an object by every time updating its coordinates table and is adequate to the robot model.
The algorithm consists of a basic process that assigns efficiently a patch of the B-reps. within a region to eight subregions when the region is divided into them. Then, this division is guided by the hierarchical structure in positioning of the octree. With the aid of both information for a region induced by the assignment and spatial information inherent in the region, the algorithm can fastly select only regions that intersect simultaneously two models.
From this selection, it follows that the interference check algorithm deals with only parts of obstacles which lie around the robot model and its computational complexity does not depend on the complexity of shape of the environment model
Finally, the computational complexity of the algorithm is evaluated, and the reasonableness of the evaluation and the efficiency of the algorithm are further ascertained by several experiments.

View full abstract

A new type sensor which can measure two dimensional environment of robot is presented. It consists of a laser diode and a CCD (Charge Coupled Device) linear array sensor built in a simple optical imaging system with a rotating mirror. Measuring range of this sensor is 250 mm to 6000 mm and measuring time for one direction is 1 ms to 10 ms. It can measure distances more than 200 points to obstacles or walls around robot per second, the position of walls is recognized from these measuring points using micro computer system.

View full abstract