Project Info
Accelerated and Heterogeneous Computing for Robotics
Neil Dantam
ndantam@mines.edu
Project Goals and Description:
Extracting full performance from modern computational devices requires effective use of a variety of specialized, heterogeneous processing units and accelerators: CPU cores, GPUs, learning accelerators, etc. Traditional techniques cannot fully utilize such heterogeneous devices. Instead, operating systems must analyze connections between computational tasks to best allocate different tasks to different processing units. In robotics applications, task dependencies further rely on interactions with the physical environment. This project develops novel programming abstractions and scheduling techniques to leverage the capabilities of modern, heterogeneous computing platforms in robotics applications.
More Information:
Grand Challenge: Engineer the tools of scientific discovery.
- SMT Introduction: https://www.cs.cmu.edu/~15811/papers/smt-intro.pdf
- SMT-LIB Tutorial: https://smtlib.github.io/jSMTLIB/SMTLIBTutorial.pdf
Primary Contacts:
Neil Dantam, ndantam@mines.edu. Mehmet Belviranli, belviranli@mines.edu.
Student Preparation
Qualifications
Student must be familiar with programming and data structures. Prior experience with programming languages and operating systems is desirable. Necessary Courses: CSCI-262. Desired Courses: CSCI-400 and CSCI-442.
TIME COMMITMENT (HRS/WK)
5
SKILLS/TECHNIQUES GAINED
Student will develop an understanding of heterogeneous computing, robot programming, and operating system scheduling. Student will learn methods and algorithms for scheduling and constraint solving.
MENTORING PLAN
Weekly individual meetings with the MURF student and faculty member to discuss progress and provide directed guidance. Weekly lab meetings to discuss the overall project and integration. Assigned graduate student mentor for the MURF student.
PREFERRED STUDENT STATUS
Junior
Senior