Software Research Seminar
- Gates Hillman 4215 and Panopto
- In Person and LiveStream ET
- TRENTON TABOR and PARV KAPOOR
► Trenton Tabor, Ph.D. Student, Software Engineering
Constraining image translation for testing: CUT-UP and other goofy Contrastive Unpaired Translation hacks
— In our ongoing work adapting image translation tools for testing autonomous vehicles, we identified and are attempting to address several characteristics of these techniques that make them ill-suited for generating robot inputs. In particular, these techniques generate images that are not necessarily: consistent across views, consistent with respect to semantics, consistent across multispectral imaging, or consistent under some function of interest. In this talk I’ll discuss our work forcing image translation to be consistent to each of these.
► Parv Kapoor, Ph.D. Student, Software Engineering
Model-based reinforcement learning from signal temporal logic specifications
— Techniques based on Reinforcement Learning (RL) are increasingly being used to design control policies for robotic systems. RL fundamentally relies on state-based reward functions to encode desired behavior of the robot and bad reward functions are prone to exploitation by the learning agent, leading to behavior that is undesirable in the best case and critically dangerous in the worst. On the other hand, designing good reward functions for complex tasks is a challenging problem. In this paper, we propose expressing desired high-level robot behavior using a formal specification language known as Signal Temporal Logic (STL) as an alternative to reward/cost functions. We use STL specifications in conjunction with model-based learning to design model predictive controllers that try to optimize the satisfaction of the STL specification over a finite time horizon. The proposed algorithm is empirically evaluated on simulations of a pick-and-place robotic arm, and other OpenAI RL environments.
In Person and LiveStream