A model-based approach to high-level robot control with visual guidance: Dissertation

Need for applying robots and robotic machines in different areas than traditional manufacturing is increasing.Much more adaptability to varying situations and uncertainties is required compared to what ordinary robot technology can offer in environments such as forests, sub-sea, or construction sites, or in very flexible production.Sensor feedback and goal-oriented behaviour are required.Guidelines on how to construct these kind of robots and machines are found from the area of intelligent robotics.This study approaches the intelligent robot control problem by viewing the computational elements and their relations in the higher levels of control hierarchy.A conceptual model with a four level hierarchical framework is proposed, consisting of three generic activities: planning, executing, and monitoring. The time dependency aspects are considered by including time constraints in plans and by describing the time relations between planning, executing, and monitoring with a specific metacontrol mechanism. -Further examination is made in three case studies.First, a human operator supported with enhanced interaction equipment is included in the robot control as a high-level task planner and environment analyzer.Second, a knowledge-based task and action planner is presented for the control of electronic component assembly with sensory guidance.Third, visual guidance for a general approach motion is formulated and an action planner is presented for planning robot motions and sensing operations of a 3D vision sensor.A special feature in the sensing in both cases of sensory guidance is the aspect of hand-eye co-operation: visual guidance is achieved with optical and vision sensors with tight integration of robot motions to the sensing process.Feasibility of the case studies is confirmed with practical and simulation experiments.