Continual Learning for Robotics: Paper Collection

A curated collection of research papers on Continual Learning (CL) for Robotics, organized by application area and searchable by technique, benchmark, and venue.

📄 Related Survey: A Survey of Continual Learning for Robotics in the Foundation Model Era

---

📖 Citation

If you find this collection useful, please consider citing our survey:

@article{Mai_2026,
    title={A Survey of Continual Learning for Robotics in the Foundation Model Era},
    url={http://dx.doi.org/10.36227/techrxiv.176972367.76460794/v2},
    DOI={10.36227/techrxiv.176972367.76460794/v2},
    publisher={Institute of Electrical and Electronics Engineers (IEEE)},
    author={Mai, Zheda and Jeon, Sooyoung and Huang, Zanming and Yoo, Jinsu and Lee, Colin and Zhang, Gengyu and Wang, Haoxuan and Li, Yifan and Kong, Yu and Yan, Yan and Chao, Wei-Lun},
    year={2026},
    month=feb
}

---

📢 News & Updates

• [2026.02.11]: Initial collection added!

---

📋 Table of Contents

• News & Updates
• How to Use This Repo
• Papers by Application
  • Manipulation
  • Navigation
  • Planning

---

🔍 How to Use This Repo

Browse by application area: Start with Manipulation, Navigation, or Planning

---

🧭 Legend

Setting

• 📦 Object
• 🎯 Goal
• 🗺️ Spatial
• 🌍 Environment

CL Technique (Abbreviations)

• GR — Generative Replay
• ER — Experience Replay
• R — Regularization
• PI — Parameter Isolation
• PEFT — PEFT
• PE — Parameter Expansion
• ME — Memory Expansion
• ML — Meta-Learning
• MoE — MoE
• W — World Model

---

📚 Papers by Application

Manipulation

Paper	CL Technique	Setting	Benchmark	Code	Venue
CRIL [Gao et al., 2021]	GR	📦🎯	Meta-World	💻	IROS’21
HyperCRL [Huang et al., 2021]	PI	🎯 / 📦🎯🗺️	Surreal Robotic Suite, DoorGym	💻	ICRA’21
HN-PPO [Schöpf et al., 2022]	R + PI	📦🎯🗺️	DoorGym	💻	NeurIPS-W’22
SANER [Powers et al., 2023]	ER + PE	📦🎯		💻	CoLLAs’23
CHN [Auddy et al., 2023]	PI	📦🎯	LASA	💻	RAS’23
CoTASP [Yang et al., 2023]	PEFT + PI	📦🎯	Continual World	💻	ICML’23
SDP [Wang et al., 2025b]	MoE + PE	📦🎯	Mimicgen, DexArt, Robomimic	💻	CoRL’24
TAIL [Liu et al., 2024]	PEFT + PE	📦 / 🎯 / 🗺️ / 📦🎯🗺️	LIBERO		ICLR’24
CompoNet [Malagon et al., 2024]	PE	📦🎯	Continual World	💻	ICML’24
LOTUS [Wan et al., 2024]	ER + PE	📦 / 🎯 / 📦🎯🗺️	LIBERO	💻	ICRA’24
ECD [Zhao et al., 2024]	ER + R	📦🎯	Continual World		ICRA’24
DMPEL [Lei et al., 2025]	PEFT + MoE + ER + PE	📦 / 🎯 / 🗺️ / 📦🎯🗺️	LIBERO	💻	ICRA’24
IsCiL [Lee et al., 2024]	PEFT + PE	📦🎯	Franka Kitchen, Meta-World	💻	NeurIPS’24
SPECI [Xu and Nie, 2025]	PE	📦 / 🎯 / 🗺️ / 📦🎯🗺️	LIBERO	💻	TCDS’25
PPL [Yao et al., 2025]	PEFT + PE	🎯 / 📦🎯	LIBERO, Mimicgen		CVPR’25
LEGION [Meng et al., 2025]	ME	📦🎯		💻	NMI’25
LLWM [Pan et al., 2025]	GR + W + R	🎯 / 📦🎯	Deepmind Control Suite, Meta-World	💻	ECML’25
iManip [Zheng et al., 2025]	PEFT + ER + PE	📦🎯	RLBench		ICCV’25
OA [Liu et al., 2025]	W + R	🎯 / 📦🎯	Continual Bench	💻	ICML’25
M2Distill [Roy et al., 2025]	R	📦 / 🎯 / 🗺️	LIBERO		ICRA’25
CKA-RL [Hu et al., 2025]	PE	🎯 / 📦🎯	Continual World	💻	NeurIPS’25
SIL-C [Lee et al., 2025]	PEFT + ME	🎯 / 📦🗺️	Franka Kitchen, Meta-World	💻	NeurIPS’25
Stellar VLA [Wu et al., 2025]	MoE + ER + ME	🎯 / 📦🎯🗺️	LIBERO		arXiv’25
TOPIC [Song et al., 2025]	PEFT + PE	📦🎯	RLBench		arXiv’25
ExpReS-VLA [Syed et al., 2025]	ER	📦 / 🎯 / 🗺️ / 📦🎯🗺️	LIBERO		arXiv’25
OMLA [Zhu et al., 2025]	PEFT + ML	📦 / 🎯 / 🗺️	LIBERO		arXiv’25
CLARE [Römer et al., 2026]	PE + PEFT	📦🎯🗺️	LIBERO	💻	arXiv’26

Navigation

Paper	CL Technique	Setting	Benchmark	Code	Venue
CVLN [Jeong et al., 2024]	ER + R	🌍	R2R, RxR		arXiv’24
VLNCL [Li et al., 2025]	ML + ER	🌍	R2R		IROS’25
UFA [Yu et al., 2025]	ER + ME	🌍	GSA-R2R		arXiv’25
TuKA [Anonymous, 2026a]	PEFT + R + PE	🌍	Habitat		ICLR’26 sub.
Uni-Walker [Anonymous, 2026b]	PEFT + R	🎯 / 🌍 / 🎯🌍	LENL		ICLR’26 sub.

Planning

Paper	CL Technique	Setting	Benchmark	Code	Venue
TAMP [Mendez-Mendez et al., 2023]	MoE + ER	📦🎯🌍	BEHAVIOR		CoRL’23
CAMA [Kim et al., 2024]	ER + R	📦🎯 / 🌍	CL-ALFRED	💻	ICLR’24
ViReSkill [Kagaya et al., 2025]	ER + ME	📦 / 🎯 / 🗺️ / 📦🎯🗺️	LIBERO		arXiv’25

---

🤝 Contributing

We welcome contributions! To add a paper:

1. Fork this repository
2. Add the paper to the appropriate application section(s)
3. Fill in all columns: Paper (with link), CL Technique, Setting, Benchmark, Code (💻 emoji if available), Venue
4. If a paper uses multiple CL techniques, list them with "+" (e.g., "PEFT + MoE + ER")
5. Submit a pull request with a brief description

Format Example:

| [Paper Title](paper-link) [Author et al., Year] | Technique(s) | Setting | Benchmark | [💻](code-link) | Venue |

---

Last Updated: February 2026 | Paper Count: 35