Joshua Powers

Papers

Total Citations

H-Index

About

Joshua Powers is a researcher at the intersection of robotics, artificial intelligence, and soft materials, whose work explores how physical form and learning algorithms interact. His primary research areas include shape-changing robotics, catastrophic forgetting in neural controllers, and the design of morphologically intelligent systems. Powers’s most impactful contribution, the 2020 paper “Morphology dictates learnability in neural controllers” (5 citations), demonstrates that a robot’s physical structure fundamentally constrains how effectively its neural controller can learn multiple tasks without catastrophic forgetting—challenging the field to consider embodiment as a key variable in machine learning. His 2021 work “Shape Matching: Evolving Fiber Constraints on a Pneumatic Bilayer” (3 citations) advances shape-changing robotics by introducing a method to program local surface curvatures in thin, stimuli-responsive sheets through differential growth and fiber constraints. Though early in his career, Powers’s research bridges artificial intelligence and soft robotics, offering a fresh perspective on how physical morphology can be leveraged to overcome learning limitations. His work is particularly notable for integrating evolutionary algorithms with material design, pointing toward robots that can adapt both their bodies and brains to changing environments.