Escaping Stability-Plasticity Dilemma in Online Continual Learning for Motion Forecasting via Synergetic Memory Rehearsal
- 1 November 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 37 (11) , 6580-6594
Abstract
Deep neural networks (DNNs) have achieved remarkable success in intelligent systems such as autonomous vehicles and robots. However, most DNN-based methods suffer from catastrophic forgetting, where the DNN may fail to maintain its performance in previously learned tasks after adapting to new data. This phenomenon gives rise to a critical tradeoff between memory stability of the DNN, that is, the ability to retain learned knowledge, and learning plasticity, that is, the capacity to acquire new information effectively. To address such stability-plasticity dilemma, this study proposes a novel CL method named synergetic memory rehearsal (SyReM). SyReM maintains a compact memory buffer to represent learned knowledge. To ensure memory stability, it employs an inequality constraint that limits increments in the average loss over the memory buffer. Synergistically, a selective memory rehearsal mechanism is designed to enhance learning plasticity by selecting samples from the memory buffer that are similar to recently observed data. This selection is based on an online-measured cosine similarity of loss gradients, ensuring targeted memory rehearsal. We validate SyReM under online CL tasks for motion forecasting. Comprehensive experiments on naturalistic driving datasets demonstrate that, compared to non-CL and CL baselines, SyReM significantly mitigates catastrophic forgetting in past scenarios while improving forecasting accuracy in new ones. The code is available at https://github.com/BIT-Jack/SyReM.Keywords
All Related Versions
Funding Information
- Key Technological Breakthrough Projects for Major Industries in Henan Province (251000240100)
- National Natural Science Foundation of China (52372405)
- Beijing Natural Science Foundation (L247033)
- China Scholarship Council
This publication has 37 references indexed in Scilit:
- MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation2025
- Flashbacks to harmonize stability and plasticity in continual learningNeural Networks, 2025
- Loss of plasticity in deep continual learningNature, 2024
- Unravelling uncertainty in trajectory prediction using a non-parametric approachTransportation Research Part C: Emerging Technologies, 2024
- Dual memory model for experience-once task-incremental lifelong learningNeural Networks, 2023
- Continual driver behaviour learning for connected vehicles and intelligent transportation systems: Framework, survey and challengesGreen Energy and Intelligent Transportation, 2023
- Biological underpinnings for lifelong learning machinesNature Machine Intelligence, 2022
- Continual learning for robotics: Definition, framework, learning strategies, opportunities and challengesInformation Fusion, 2020
- Overcoming catastrophic forgetting in neural networksProceedings of the National Academy of Sciences of the United States of America, 2017
- Catastrophic Interference in Connectionist Networks: The Sequential Learning Problem1989