The core premise of this research is to enhance physical crowdsourcing by designing interactions, algorithms, and architectures that produce globally effective behaviors from coordinated opportunistic contributions. Our research aims to enable physical crowdsourcing systems to tap into the daily routines of over 100 million on-the-go Americans to transport goods, map the world in exquisite new detail, and accomplish a broad range of tasks at scale. We propose a general framework and set of techniques that aim to achieve this by (1) scaffolding individual contributions toward a communal goal, (2) developing computational mechanisms that flexibly guide people to appropriate tasks and intelligently manage community participation, and (3) evaluating the effectiveness of these techniques through empirical simulations and field-based user evaluations. Recent years have brought about physical crowdsourcing systems that help connect people to tasks and lower barriers to participation. Communitysensing projects recruit volunteers to track invasive species, report city infrastructure problems, and tag geographic points of interest. Yet, physical crowdsourcing applications are often limited by the fact that contributions are gathered via one of two approaches: opportunistic or directed. Opportunistic approaches rely on input from users in situations where users decide to contribute; systems cannot make demands for particular data to be collected or tasks to be completed, making it difficult to guarantee high-resolution data or ensure timely task completion to maintain quality of service. Directed approaches prompt users for input that is beneficial to achieving high-level system goals. However, directed approaches often require contributions outside people's existing routines and as a result strong incentives are required. We propose a hybrid approach to physical crowdsourcing that brings about the best elements of both opportunistic and directed approaches. Our approach permits access to a large number of people who can contribute through their routines in ways that better achieve desired system goals, such as: (1) collecting high fidelity physical data, expanding data coverage, and filling gaps in existing data coverage; (2) completing physical tasks efficiently and with minimal disruption along individual helpers' existing routes; and (3) achieving desired quality of service with minimal disruption to the routines of our crowd to promote efficiency and ensure long-term system health. Intellectual Merit This work builds theory, algorithms and frameworks for indirectly coordinating large numbers of autonomous individuals, who when induced to make small, convenient contributions through their existing routines, will volunteer actions and data to achieve complex system goals. The work will enable new opportunities for physical crowdsourcing over large geographic areas. The research also advances the study of opportunistic planning by introducing a new framework that enables flexible, goal-directed approaches for coordinating opportunistic actions in ways largely absent in the physical domain. Our work addresses a need for methods that can plan and execute collective actions opportunistically based on mobility resources, and that can make adjustments over the course of problem solving based on situations on the ground. The research is of broad interest to crowdsourcing and artificial intelligence researchers, HCI researchers, cognitive scientists, and engineers. Broader Impact This research will contribute directly to several societa
|Effective start/end date||7/1/16 → 6/30/21|
- National Science Foundation (IIS-1618096)
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.