Abstract:
This paper presents an adaptive model-free deep reinforcement approach that can recognize and adapt to ride-sharing diurnal patterns with car-pooling. Deep Reinforcement Learning (RL) suffers from catastrophic forgetting because it is agnostic to experience distribution timescales. RL algorithms always converge to optimal policies in Markov decision processes (MDPs) in static environments. This assumption is restrictive. RL methods produce suboptimal decisions in highly dynamic environments like ride-sharing, traffic control, etc. In highly dynamic environments, we (1) use an online Dirichlet change point detection (ODCP) algorithm to detect changes in experience distribution and (2) develop a Deep Q Network (DQN) agent that can recognize diurnal patterns and make informed dispatching decisions based on the underlying environment. Instead of fixing patterns by time of week, the proposed approach automatically detects that the MDP has changed and uses the new model. This paper proposes a dynamic, demand-aware vehicle-passenger matching and route planning framework based on online demand, vehicle capacities, and locations. Our approach improves fleet utilization by using less than 50% of the fleet to serve up to 90% of requests while maximizing profits and minimizing idle time, as shown by the New York City Taxi public dataset.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
