A variety of animals such as primates, dogs, and bears switch modes of locomotion between quadrupedalism and bipedalism to better complete certain tasks. However, very few robotic platforms can effectively combine the two forms of locomotion. A multi-modal robotic platform with such capabilities would provide additional adaptability in unstructured environments, broadening its potential applications. Therefore, we extended an existing quadrupedal platform with the capability to transition into a bipedal stance. In this project, we built a physical robot, developed an accompanying software stack with a reinforcement learning pipeline, implemented quadrupedal locomotion, and achieved stance transition in simulation. Our integrated hardware and software platform affords future roboticists the opportunity to test and develop more adaptable locomotion strategies and increase the functionality of robots more broadly.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Euredjian, Andrew G.
• Mahajan, Revant
• Gupta, Ankur

Subject

• Learning
• Computing
• Project-Based Learning

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-051021-141909
• 23751

Mot-clé

• quadrupedal
• multi-modal robot
• transforming robot
• bipedal
• reinforcement learning
• simulation

Advisor

• Agheli Hajiabadi, Mohammad Mahdi
• Paffenroth, Randy Clinton
• Gennert, Michael A.

Year

• 2021

Date created

• 2021-05-10

Resource type

• Major Qualifying Project

Major

• Robotics Engineering
• Computer Science

Rights statement

• In Copyright

License

• Creative Commons BY Attribution 4.0 International

Dernière modification

• 2021-09-15