AI and Robotics Wheeler Ruml Department of Computer Science - - PowerPoint PPT Presentation

Wheeler Ruml (UNH) AI and Robotics – 1 / 5

AI and Robotics

Wheeler Ruml Department of Computer Science www.cs.unh.edu/~ruml

Research

■ Research ■ Integration ■ Challenges

Wheeler Ruml (UNH) AI and Robotics – 2 / 5

Planning: ideas from combinatorial search

■

canon (Robin): models (abstraction) + inference (search)

■

abstraction-based guidance

■

suboptimal search (bounded, contract, utility-based)

■

real-time planning, concurrent execution (Manuela)

■

multi-level planning (Max, Dawn): hierarchical search

Research

■ Research ■ Integration ■ Challenges

Wheeler Ruml (UNH) AI and Robotics – 2 / 5

Planning: ideas from combinatorial search

■

canon (Robin): models (abstraction) + inference (search)

■

abstraction-based guidance

■

suboptimal search (bounded, contract, utility-based)

■

real-time planning, concurrent execution (Manuela)

■

multi-level planning (Max, Dawn): hierarchical search Perception:

■

focus of attention, ‘purposeful’ (Jana, Manuela, Matthias)

Encouraging Integration

■ Research ■ Integration ■ Challenges

Wheeler Ruml (UNH) AI and Robotics – 3 / 5

Funders:

■

fund workshop invited speakers

■

fund presentation of journal papers at related conferences

■

fund workshops that rotate among conferences

■

fund summer school or workshop (Robin)

■

how to encourage publishing in diverse venues?

Encouraging Integration

■ Research ■ Integration ■ Challenges

Wheeler Ruml (UNH) AI and Robotics – 3 / 5

Funders:

■

fund workshop invited speakers

■

fund presentation of journal papers at related conferences

■

fund workshops that rotate among conferences

■

fund summer school or workshop (Robin)

■

how to encourage publishing in diverse venues? Event organizers:

■

umbrella model: present best papers from related conferences, ‘what’s hot’ talks

■

clearly demarcate outreach from research events, invite

• rganizers to report out on current technical approaches and

challenges

■

( not co-location: too hard )

Challenge Problems

■ Research ■ Integration ■ Challenges

Wheeler Ruml (UNH) AI and Robotics – 4 / 5

■

cheap reliable common platform (UBR-1)

■

works in simulation

■

min ∆ ‘shaping’. eg: Physical TSP, setting the table

■

ask people with jobs, money, cool factor (≈ David E. Smith)

■

counter-balance with explicit encouragement of exploratory research on new problem settings (Tomas)