| MODELING MOTORCYCLE MANEUVERING IN URBAN SCENARIOS USING MARKOV DECISION PROCESS WITH A DYNAMICAL-DISCRETIZED REWARD FIELD |
| Rina Mardiati1,3, Bambang Riyanto Trilaksono1, Sony Sulaksono Wibowo1, Dina Shona Laila2 |
1Bandung Institute of Technology
2Coventry University
3UIN Sunan Gunung Djati Bandung |
|
|
|
|
|
|
|
| ABSTRACT |
|
This paper proposes a novel MDP framework to deal with the accuracy of the motorcycle driving model. It proposes a weighted and unweighted Dynamical-Discretized Reward Field (DDRF) as a major contribution on modeling motorcycle maneuver in mixed traffic conditions. Other contributions of this work are the integration of a motorcycle trajectory maneuver model in the state transition function, derivation of probability functions, area of awareness (AoA) and its sectorization to perceive vehicles inside the AoA, which is used to determine actions. We conducted some simulations to evaluate the performance of the proposed model by comparing the data from the simulations with real data. In this study, we use 100 simulation data on motorcycle maneuvering, which consisted of two different scenarios, i.e., 50 data of motorcycle maneuvering to avoid other motorcycles and 50 data of motorcycle maneuvering to avoid cars. We adjusted the simulation setting to the real situation and measured the performance of the proposed model using root mean square error (RMSE). In general, the proposed method can properly model the maneuver of motorcycles in heterogeneous traffic with an RMSE value of around 0.74 meters. This model performs twice as good as an existing car-following model. Furthermore, the proposed reward function performs around 4 ~ 6 % better than the reward function in previous studies. |
| Key Words:
Markov decision process, Motorcycle, Reward function, Driving behavior model |
|
|
|
Corresponding Author.
Bambang Riyanto Trilaksono , Email. 
Preview
Full text via DOI
Download Citation 
Print 
