Probabilistic Robotics 第二章总结.docxVIP

Key Topic 1:Robot Environment InteractionStateEnvironments are characterized by state.State State at time t Typical State VariablesRobot pose: 6 state variables: (x, y, z) for Cartesian coordinates; (pitch, roll, yaw) for angular orientation (Kinematic State)Configuration of the robots actuators. (Kinematic State)Robot velocity and the velocities of its joints. 6 variables, each for 1 pose variable. (Dynamic State)Location and features of surrounding objects in the environment.Location and velocities of moving objects and people.A huge number of others.Complete StateBest predictor of the future.Completeness entails that knowledge of past states, measurements or controls carry no additional information that would help us to predict the future more accurately.Incomplete StateIn practice, it is impossible to specify a complete state for any realistic robot system.Therefore, practical implementations single out a small subset of all state variables. IncompletenessHybrid State SpaceState can be continuous (e.g., robot pose), can be discrete (e.g., sensor broken or not).State spaces that contain both continuous and discrete variables are called Hybrid state spaces.Environment InteractionTwo fundamental types of interaction between a robot and its environment:Robot influence the state of Environment (Via actuator)Control actions.Always execute a control action, even when no motor is moved.Robot gather state info from Environment (Via sensor)Percept, observation or measurement.Always have delays.Two different data streams:Measurement data:Control data:Probabilistic Generative LawsThe probabilistic law characterizing the evolution of state:depending on all past states, measurements and controls.If state is complete,State TransitionMeasurement ProbabilityBelief DistributionsA belief reflects the robot’s internal knowledge about the state of the environment. A belief, or state of knowledge with regards to a state should be distinguished from the true state itself.We den