This paper presents an octree algorithm for collision and interference detection using space partitioning technique. The technique greatly reduces the computation time consumed in dynamic collision detection during simulation progress. The simulated objects are represented in hierarchically decomposed octrees. Under this technique, the checking space can be partitioned according to the geometric dependence of two octrees, so that the relations (overlap or separate) among the nodes in the octrees can be determined directly. Since heuristic calculation is excluded from the algorithm, the time consumption for collision detection is greatly reduced.