Multi-task imitation learning (MTIL) has shown significant potential in robotic manipulation by enabling agents to perform various tasks using a unified policy. This simplifies the policy deployment and enhances the agent's adaptability across different contexts. However, key challenges remain, such as maintaining action reliability (e.g., avoiding abnormal action sequences that deviate from nominal task trajectories), distinguishing between similar tasks, and generalizing to unseen scenarios. To address these challenges, we introduce the Foresight-Augmented Manipulation Policy (FoAM), an innovative MTIL framework. FoAM not only learns to mimic expert actions but also predicts the visual outcomes of those actions to enhance decision-making. Additionally, it integrates multi-modal goal inputs, such as visual and language prompts, overcoming the limitations of single-conditioned policies. We evaluated FoAM across over 100 tasks in both simulation and real-world settings, demonstrating that it significantly improves IL policy performance, outperforming current state-of-the-art IL baselines by up to 41% in success rate. Furthermore, we released a simulation benchmark for robotic manipulation, featuring 10 task suites and over 80 challenging tasks designed for multi-task policy training and evaluation.

@misc{liu2024foamforesightaugmentedmultitaskimitation,
        title={FoAM: Foresight-Augmented Multi-Task Imitation Policy for Robotic Manipulation},
        author={Litao Liu and Wentao Wang and Yifan Han and Zhuoli Xie and Pengfei Yi and Junyan Li and Yi Qin and Wenzhao Lian},
        year={2024},
        eprint={2409.19528},
        archivePrefix={arXiv},
        primaryClass={cs.RO},
        url={https://arxiv.org/abs/2409.19528},
      }

Robotic manipulation continues to be a complex challenge, with imitation learning (IL) offering an effective way for robots to learn tasks from expert demonstrations. Current IL methods typically rely on fixed camera setups—either multi-camera systems, which may introduce redundant or noisy data, or single-camera systems, which suffer from limited viewpoints, constraining task performance. Inspired by human active perception, where humans dynamically adjust their viewpoint to capture the most relevant and least noisy information, we propose MAE-Select, a novel framework for active viewpoint selection in single-camera robotic systems. MAE-Select fully leverages pre-trained multi-view masked autoencoder representations and dynamically selects the next most informative viewpoint at each time chunk without requiring labeled viewpoints. This plug-and-play approach enhances learning efficiency and task performance. Extensive experiments demonstrate that MAE-Select improves the capabilities of single-camera systems and, in some cases, even surpasses multi-camera setups.

    @misc{zhai2025visionlanguageactioncriticmodelroboticrealworld,
        title={A Vision-Language-Action-Critic Model for Robotic Real-World Reinforcement Learning}, 
        author={Shaopeng Zhai and Qi Zhang and Tianyi Zhang and Fuxian Huang and Haoran Zhang and Ming Zhou and Shengzhe Zhang and Litao Liu and Sixu Lin and Jiangmiao Pang},
        year={2025},
        eprint={2509.15937},
        archivePrefix={arXiv},
        primaryClass={cs.RO},
        url={https://arxiv.org/abs/2509.15937}, 
    }
    

Robotic manipulation continues to be a complex challenge, with imitation learning (IL) offering an effective way for robots to learn tasks from expert demonstrations. Current IL methods typically rely on fixed camera setups—either multi-camera systems, which may introduce redundant or noisy data, or single-camera systems, which suffer from limited viewpoints, constraining task performance. Inspired by human active perception, where humans dynamically adjust their viewpoint to capture the most relevant and least noisy information, we propose MAE-Select, a novel framework for active viewpoint selection in single-camera robotic systems. MAE-Select fully leverages pre-trained multi-view masked autoencoder representations and dynamically selects the next most informative viewpoint at each time chunk without requiring labeled viewpoints. This plug-and-play approach enhances learning efficiency and task performance. Extensive experiments demonstrate that MAE-Select improves the capabilities of single-camera systems and, in some cases, even surpasses multi-camera setups.

Heat exchangers (HX) have a wide range of applications in various fields, such as industrial production and aerospace. The rapid development of additive manufacturing (AM) in recent years has made it possible to produce HXs with complex internal structures. Among these, the HX with triply periodic minimal surface (TPMS) structures has drawn the attention of many researchers due to its large area-to-volume ratio, flexible design capabilities, and the feature of separating hot and cold flow channels. It has been believed to have the potential to replace traditional plate and tube HXs and revolutionize the entire field of HX design. TPMS is a specific type of three-dimensional surface with certain mathematical formula, and is suitable for constructing high performance HX. Among the TPMS structures, the F-RD topology shows good mechanical properties, while the flow characteristics and mechanism inside F-RD HX have not been clarified yet. Therefore, this study reports the design and fabrication of an F-RD aluminum HX via AM. By conducting numerical simulations and heat transfer experiments, the internal flow characteristics and heat transfer performance of the F-RD HX are analyzed. Based on the literature review, the S-D topology exhibits the best heat transfer performance among TPMS structures. This research findings show that the F-RD topology demonstrates comparable heat transfer performance in contrast with the SD topology. However, due to the different blockage areas, the F-RD HX exhibits different pressure drops in hot the and cold channels, which are larger than those of the S-D HX. These characteristics allow F-RD HXs to be the suitable option for scenarios that require both good mechanical and heat transfer performance. In addition, to improve the design flexibility of HXs and adapt to a wider range of application scenarios, this study also explores the modeling method of Hybrid-TPMS structure. The findings of this study will provide valuable insights for the optimal design of HXs.

The crash of an aircraft with an almost vertical attitude in Wuzhou, Guangxi, China, on 21 March 2022, has caused a robust discussion in the civil aviation community. We propose an active disturbance rejection controller (ADRC) for suppressing aeroelastic vibrations of a flexible aircraft at the simulation level. The ADRC has a relatively simple structure and it has been proved in several fields to provide better control than the classical proportional-integral-derivative (PID) control theory and is easier to translate from theory to practice compared with other modern control theories. In this paper, the vibration model of the flexible aircraft was built, based on the first elastic vibration mode of the aircraft. In addition, the principle of ADRC is explained in detail, a second-order ADRC was designed to control the vibration model, and the system’s closed-loop frequency domain characteristics, tracking effect and sensitivity were comprehensively analyzed. The estimation error of the extended state observer (ESO) and the anti-disturbance effect were analyzed, while the robustness of the closed-loop system was verified using the Monte Carlo method, which was used for the first time in this field. Simulation results showed that the ADRC suppressed aircraft elastic vibration better than PID controllers and that the closed-loop system was robust in the face of dynamic parameters.

@article{liu2022comprehensive,
  title={Comprehensive Engineering Frequency Domain Analysis and Vibration Suppression of Flexible Aircraft Based on Active Disturbance Rejection Controller},
  author={Liu, Litao and Tian, Bingwei},
  journal={Sensors},
  volume={22},
  number={16},
  pages={6207},
  year={2022},
  publisher={MDPI}
}
                                              

Gold and Bitcoin have long been popular among investors as traditional and newer assets that can fight inflation and preserve value, respectively, which has led to an increasing number of investors interested in investing in Bitcoin and gold. This paper focuses on a trading decision model based on time series analysis combined with innovative bull and bear market forecasts. By comparing the forecasting effectiveness of Autoregressive Integrated Moving Average (ARIMA) and XGBoost neural network, we decided to choose the more effective ARIMA model for stock price forecasting. ARIMA models combined with bull and bear market forecasting models for conservative stock traders analyze the risk and return of trading and propose trading decisions. In addition to this, we verified the optimality of the developed model and analyzed the sensitivity of the model. Adjustments were made to the gold, bitcoin and cash positions on 23 different dates through the trading decision methodology we provided. Ultimately, the model we built is expected to deliver around 260 times profitability over the dates of the data used.

@inproceedings{liu2022stock,
  title={Stock Investment and Trading Strategy Model Based on Autoregressive Integrated Moving Average},
  author={Liu, Litao},
  booktitle={2022 IEEE Conference on Telecommunications, Optics and Computer Science (TOCS)},
  pages={732--736},
  year={2022},
  organization={IEEE}
}
                                              

In recent years, China has adhered to the concept of sustainable development of respecting nature, protecting nature and living in harmony with nature. A typical example is China’s long-term ecological management of Saihanba. Since 1962, after more than half a century of rescue, China has established the world’s largest artificial forest, improving the ecological environment of Saihanba area, making Saihanba the lungs of surrounding cities (Beijing, Tianjin, etc.). We collect relevant environmental data from 2002 to 2020, established 11 indicators through principal component analysis, score the annual environmental quality of Saihanba with 99% confidence, and propose an environmental assessment model for Saihanba. In second part, we collect environmental data in Beijing in recent years, such as the frequency of sandstorms, particle concentration, etc., and aims to explore the benefits of Saihanba governance to surrounding cities. The results show that after the implementation of environmental management measures, the environmental quality of Saihanba has been improved, and it has had a positive impact on the surrounding cities, that is, improved the environmental quality of nearby cities such as the Beijing and Tianjin.

@inproceedings{liu2022research,
  title={Research on Ecological Evaluation Model Based on Principal Component Analysis},
  author={LIU, Litao and LI, Lei and MO, Haozhou},
  booktitle={Proceedings of the 3rd International Conference on Green Energy, Environment and Sustainable Development (GEESD2022)},
  volume={23},
  pages={429},
  year={2022},
  organization={IOS Press}
}
                                              

Disasters can have a significant impact on air quality and thus pose a threat to the health of people in the affected areas. There are no post-disaster air purification systems or specialized technologies currently. In this paper, a post-disaster air purification system is designed based on the mechanism of lung purification of mammals. It uses two-stage purification, which mainly purifies particulate matter in the air, such as PM2.5 and PM10. The first stage is coarse purification, which simulates the villi fibers in the nasal cavity to block particulate matter. The second stage simulates macrophages in the alveoli to achieve efficient purification of particulate matter. The post-disaster air purification system adopts intelligent control and introduces an innovative mechanical structure design with a special adsorption structure and guiding structure, which can effectively increase the adsorption area of particulate matter in the air and its adsorption capacity. Also, it has a simple structure and can be disassembled and assembled with less effort, which meets the special emergency needs after the disaster.

@inproceedings{liang2022post,
  title={Post-disaster air purification system based on bionic lung design},
  author={Liang, Yonggang and Ling, Yuewei and Li, Yujie and Liu, Litao and Xu, Nuo},
  booktitle={International Conference on Automation Control, Algorithm, and Intelligent Bionics (ACAIB 2022)},
  volume={12253},
  pages={462--467},
  year={2022},
  organization={SPIE}
}