Selected work

Representative Papers

Beyond Pure Sampling: Hybrid Optimization Mechanisms for Non-Convex Model Predictive Control

May 30, 2026

This work addresses the challenge of navigating complex cost landscapes in non-convex model predictive control, where nonlinear dynamics and multiple obstacles often trap gradient-based methods in suboptimal local minima. To overcome this limitation, we propose a Maximum Entropy Differential Dynamic Programming (ME-DDP) framework that integrates deterministic optimization with entropy-maximizing sampling. Our approach employs a two-stage mechanism: it first performs local gradient-based refinement via DDP and then leverages the inverse Hessian of the action-value function to guide policy sampling, enabling escape from local minima and balancing global exploration with local exploitation. We develop three ME-DDP variants, elucidate their theoretical connections to Model Predictive Path Integral (MPPI) control, and demonstrate superior performance across four navigation benchmarks—achieving higher success rates in high-dimensional systems, outperforming MPPI in low-dimensional settings, and exhibiting robustness in real-world quadrotor experiments through dense obstacle fields.

0 citationsRead paper

FORWARD: Dataset of a forwarder operating in rough terrain

Nov 21, 2025

To address key challenges in terrain traversability assessment, environmental perception, and autonomous control of forestry machinery operating in rugged terrain, this study introduces the first high-precision, multimodal dataset specifically designed for timber extraction operations in forested environments. Leveraging synchronized data acquisition from RTK-GNSS, 360° panoramic vision, IMU, CAN bus, vibration sensors, and high-density LiDAR-based terrain scanning, we collected 18 hours of real-world operational data across diverse Swedish forest sites—covering varying payload conditions, vehicle speeds, and track configurations. The dataset provides centimeter-level georeferencing, standardized Stanford-style logging formats, and fine-grained annotations of operational primitives. It represents the first long-duration, multi-scenario, high spatiotemporal-resolution record of end-to-end forestry machinery operations, enabling advances in traversability modeling, energy consumption optimization, autonomous navigation algorithm development, and digital twin simulation. This resource establishes a foundational data infrastructure and methodological framework for intelligent forestry equipment research and development.

0 citationsRead paper

Robotics Under Construction: Challenges on Job Sites

Jun 24, 2025

To address the dual challenges of labor shortages and stagnant productivity in the construction industry, this paper proposes and validates an autonomous material transportation system tailored for dynamic construction environments. Methodologically, the system is built upon the CD110R-3 tracked platform and integrates high-precision GNSS positioning, multi-source external sensor fusion for perception and mapping, adaptive terrain navigation, and multi-robot cooperative scheduling—thereby overcoming critical bottlenecks including environmental dynamism, evolving terrain topology, and suboptimal sensor placement. Experimental validation in real-world construction sites demonstrates the system’s feasibility and robustness, significantly enhancing environmental adaptability and operational efficiency of unmanned material transport. This work provides a deployable technical pathway and foundational insights for robotics-enabled intelligent construction, identifying autonomous navigation, construction-aware perception, and swarm-level coordination as three core research directions.

0 citationsRead paper

Machine Learning-Based Self-Localization Using Internal Sensors for Automating Bulldozers

Jun 08, 2025

To address autonomous bulldozer localization failure in RTK-GNSS-denied environments—common in mining operations—this paper proposes a satellite-independent, multi-sensor fusion localization method. First, we construct the first bulldozer-specific odometry dataset, incorporating operation-critical sensors such as blade pose and hydraulic pressure. Second, we employ LSTM/MLP models to fuse IMU, wheel-speed, and hydraulic measurements for robust local velocity estimation. Finally, an Extended Kalman Filter (EKF) recursively integrates these estimates to compute global pose. The method significantly enhances robustness under challenging conditions—including wheel slip, steep inclines, and excavation—where conventional kinematic models degrade rapidly. In serpentine driving and other representative scenarios, positional drift is reduced by 42% compared to traditional approaches. Experimental validation confirms both the feasibility and superiority of this GNSS-free solution for real-world mining applications.

0 citationsRead paper

Recent publications

Latest Papers

Beyond Pure Sampling: Hybrid Optimization Mechanisms for Non-Convex Model Predictive Control

May 30, 2026

This work addresses the challenge of navigating complex cost landscapes in non-convex model predictive control, where nonlinear dynamics and multiple obstacles often trap gradient-based methods in suboptimal local minima. To overcome this limitation, we propose a Maximum Entropy Differential Dynamic Programming (ME-DDP) framework that integrates deterministic optimization with entropy-maximizing sampling. Our approach employs a two-stage mechanism: it first performs local gradient-based refinement via DDP and then leverages the inverse Hessian of the action-value function to guide policy sampling, enabling escape from local minima and balancing global exploration with local exploitation. We develop three ME-DDP variants, elucidate their theoretical connections to Model Predictive Path Integral (MPPI) control, and demonstrate superior performance across four navigation benchmarks—achieving higher success rates in high-dimensional systems, outperforming MPPI in low-dimensional settings, and exhibiting robustness in real-world quadrotor experiments through dense obstacle fields.

0 citationsRead paper

FORWARD: Dataset of a forwarder operating in rough terrain

Nov 21, 2025

To address key challenges in terrain traversability assessment, environmental perception, and autonomous control of forestry machinery operating in rugged terrain, this study introduces the first high-precision, multimodal dataset specifically designed for timber extraction operations in forested environments. Leveraging synchronized data acquisition from RTK-GNSS, 360° panoramic vision, IMU, CAN bus, vibration sensors, and high-density LiDAR-based terrain scanning, we collected 18 hours of real-world operational data across diverse Swedish forest sites—covering varying payload conditions, vehicle speeds, and track configurations. The dataset provides centimeter-level georeferencing, standardized Stanford-style logging formats, and fine-grained annotations of operational primitives. It represents the first long-duration, multi-scenario, high spatiotemporal-resolution record of end-to-end forestry machinery operations, enabling advances in traversability modeling, energy consumption optimization, autonomous navigation algorithm development, and digital twin simulation. This resource establishes a foundational data infrastructure and methodological framework for intelligent forestry equipment research and development.

0 citationsRead paper

Robotics Under Construction: Challenges on Job Sites

Jun 24, 2025

To address the dual challenges of labor shortages and stagnant productivity in the construction industry, this paper proposes and validates an autonomous material transportation system tailored for dynamic construction environments. Methodologically, the system is built upon the CD110R-3 tracked platform and integrates high-precision GNSS positioning, multi-source external sensor fusion for perception and mapping, adaptive terrain navigation, and multi-robot cooperative scheduling—thereby overcoming critical bottlenecks including environmental dynamism, evolving terrain topology, and suboptimal sensor placement. Experimental validation in real-world construction sites demonstrates the system’s feasibility and robustness, significantly enhancing environmental adaptability and operational efficiency of unmanned material transport. This work provides a deployable technical pathway and foundational insights for robotics-enabled intelligent construction, identifying autonomous navigation, construction-aware perception, and swarm-level coordination as three core research directions.

0 citationsRead paper

Machine Learning-Based Self-Localization Using Internal Sensors for Automating Bulldozers

Jun 08, 2025

To address autonomous bulldozer localization failure in RTK-GNSS-denied environments—common in mining operations—this paper proposes a satellite-independent, multi-sensor fusion localization method. First, we construct the first bulldozer-specific odometry dataset, incorporating operation-critical sensors such as blade pose and hydraulic pressure. Second, we employ LSTM/MLP models to fuse IMU, wheel-speed, and hydraulic measurements for robust local velocity estimation. Finally, an Extended Kalman Filter (EKF) recursively integrates these estimates to compute global pose. The method significantly enhances robustness under challenging conditions—including wheel slip, steep inclines, and excavation—where conventional kinematic models degrade rapidly. In serpentine driving and other representative scenarios, positional drift is reduced by 42% compared to traditional approaches. Experimental validation confirms both the feasibility and superiority of this GNSS-free solution for real-world mining applications.

0 citationsRead paper