"Probabilistic robot" learning note one

Introduction to the first chapter

1.1 Uncertainties in robotics

Robots must be able to accept a large number of uncertainties in the objective world. There are many factors that cause the robot to be uncertain:

Robot environment

Sensor

Actuator of the robot

Robot software

Approximate algorithm

1.2 Probabilistic Robotics

Probabilistic Robotics focuses on the uncertainty of robot perception and behavior. The main idea of probabilistic robot is to use the calculation of probability theory to express the uncertainty clearly.

1.3 Revelation

The probabilistic robot integrates the model seamlessly with the sensor data and overcomes the limitations of both.

Compared with the traditional model-based robot technology, the probabilistic algorithm has lower accuracy requirements for the robot model. The probabilistic algorithm is less accurate than many reaction techniques, and its only control input is the input of the instantaneous sensor.

From the perspective of probability, the robot learning problem (robotic learning problem) is a long-term estimation problem.

Probabilistic algorithm limitations: Computational complexity, approximation of inevitability.