PAL 9000

PAL, Prismatic Automated Locomotive, is the result of a research effort investigating the application of prismatic structures as informal robotic armatures. The project identifies prismatic structures as open, multi-faceted surfaces which are derived through the extrusion of selected faces of a base polyhedron. By identifying prismatic structures which possess one or more degrees of freedom, we are then able to reorient, reconfigure, array and actuate these mechanisms to produce robotically useful movement. ‘Soft’ materials such as cardstock and thin chipboard are utilized to construct prototypes, relating the work to folded tectonics and the notion of informal robotics, whereby rigid, hard materials are replaced by the relatively pliant. Lastly, PAL is conceived of as an interactive component of a complex, whereby a number of agents are able to demonstrate a variety of behaviors within an automated system.

PAL’s control system utilizes an overhead camera for global positioning. Each robot has a uniquely colored isosceles triangle on its top surface. The trajectory of the robot is identified by calculating the longest line from the centroid of the triangle to each end point. Using computer vision, we developed an algorithm that would identify our robotic ‘seekers’ and their ‘prizes.’ Each robotic ‘seeker’ would find the nearest prize of its own color and move towards it, ignoring others.

Project completed for Informal Robotics, Harvard Graduate School of Design with Chuck Hoberman

Team.  Kate Hajash (SMArchS, MIT) + Joe Varholick (MARCHII, GSD)
Role. Prototyped early design mechanisms. Developed framework for sensing and communication. Programmed and developed control system (Python) using an overhead camera, computer vision (OpenCV), python communication, and Arduino.

Evolution of Design + Mechanism

evolution of control systems

The final control system is guided by an overhead camera, mounted above the testing area. Using computer vision, we developed an algorithm that could complete shape and color recognition. Seekers were marked with an isosceles triangle, while the prizes were rectangles. There were three different colors: red, green and blue. Each seeker could only search for its own color. The control system would identify each seekers closest color and calculate the angle at which the robot needed to turn. Using PySerial, we sent our instructions to an Arduino connected to the computer. This Arduino then sent instructions through radio frequency to each remote robot.