Robotic Automation Using PyBullet

Robotic Automation Using PyBullet

Dhanesh Amin1, Nihar Kadam2, Shubhangi Mahadik3

1 Student, MCA Bharti Vidyapeeth Navi Mumbai, India

2 Student, MCA Bharti Vidyapeeth Navi Mumbai, India

3 Asst. Professor, MCA Bharti Vidyapeeth

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - Physics-based simulations of passive phenomena such as substances and liquids have become almost ubiquitous in the industry. However, the introduction of physically simulated characters was more modest. Modelling the movements of humans and animals remains a difficult problem, and there are currently few ways to simulate the different behaviours shown in the real world. Permanent challenges in this area include generalization and manageability. While manually designed controller-based methods have produced compelling results, the availability of human insights limits new skills and the ability to generalize to new situations. Humans can perform a variety of skills themselves, but it is difficult to clarify the internal strategies underlying such skills, and it is even more difficult to code them into a controller. Mobility is another obstacle that has hindered the adoption of simulated characters. Creating motion for a simulated character is still notorious for being difficult, and the current interface does not provide users with an effective means of deriving the desired movement from the simulated character. Reinforcement learning offers a promising approach to motion synthesis. This approach reduces the need for human insight by learning that agents perform different skills through trial and error.

Key Words:  Physics-based simulation, Reinforcement Learning, motion synthesis, Simulated characters