Smart Return System: Optimizing Bike Return Experience Using Sensor Data and Trajectory Prediction

This article discusses the challenges and solutions in optimizing the bike return process for shared mobility services. The current system faces issues such as poor user experience, high risk factors, slow return speeds, complex processes, and inaccurate location detection. To address these problems, the article proposes a smart return system that leverages sensor data and trajectory prediction.

The system utilizes accelerometer data from the bike's intelligent components to predict return behavior and perform early judgment. By analyzing the acceleration patterns, the system can distinguish between temporary stops (like waiting at traffic lights) and actual parking. This allows for 85% of orders to be pre-judged before the user clicks 'return'.

Additionally, the system employs trajectory prediction using Hidden Markov Models to correct the bike's path based on road network information. This helps in accurately determining return locations even when the user's phone location is unreliable or delayed.

The smart return system significantly improves the user experience by reducing return time and increasing accuracy. It also helps in reducing operational costs and improving overall service quality for the shared mobility provider.