Probabilistic path planning for a six degree-of-freedom articulated robot in an industrial dynamic environment using simultaneous localization and mapping (SLAM) [EMBARGOED]

Abstract

An industrial robot is typically isolated into work cells, rendering the operation of in- dustrial robots rigid and inflexible. Should a new task or change be required in the work cell, the robot demands a tremendous amount of time and resources. This reality of- ten hinders the confidence of Small and Medium-Sized Enterprise (SME) in adopting industrial robot technology. This work aims to provide a solution to automate SMEs by introducing flexible and safe robotic technology. The research demonstrate the fea- sibility of constructing a robot manipulator with obstacle avoidance motion planning. Therefore, a manipulator prototype is developed to investigate the planning and motion control of the manipulator in an industrial dynamic environment. Sampling-based plan- ners are tested with dynamic and static but unpredictable obstacles of varying shapes. Rapidly-Exploring Random Tree (RRT) is chosen to plan and control the motion of the manipulator in the dynamic environment. A physics engine simulator is used to sim- ulate the manipulator and obstacle avoidance planning. The experimentation on static unpredictable obstacles reveals that the cycle space generator algorithm invokes reac- tive behavior. Furthermore, 11 planners successfully complete 100 planning requests with 50 cycle space iterations without failure during the simulation and experiment. For the experimentation involving the dynamic obstacle, a three-dimensional (3D) obstacle is introduced into the simulation environment and moves periodically from an initial point to a goal point in the form of sinusoidal motion described by A + B sin (2?tC). However, the moving obstacle is augmented by the simulation environment to ensure safe experimentation and preserve the prototype. It is observed that the global plan- ner, RRT, behaves reactively when subjected to the cycle space defined in this report. Consequently, Real-Time Appearance-Based Mapping (RTAB-Map) is used to facilitate encoder-less context experimentation. RTAB-Map, as a Simultaneous Localization and Mapping (SLAM) solution, provides confident state estimation of the robot task space and, consequently, the joint-space configuration of the manipulator. Nonetheless, the pose-graph optimizations in RTAB-Map’s visual odometry failed to produce pose esti- mates error of less than 0.001 m for position estimate and 0.01 for rotation estimate. This prevents further investigation on an encoder-less system within the scope of this research. It is recommended that further investigation on the encoderless system should use Oriented FAST and ROTATED BRIEF descriptor (ORB) solutions to materialize a Planner-SLAM solution.