Published papers

本論文では，日本のアストロバイオロジーの研究者グループはJAMP（Japan Astrobiology Mars Project）と呼ばれる火星での生命探査計画を紹介し，これまでの生命関連探査とJAMP計画を概説するとともに，火星生命探査に用いる3 つの機器（蛍光顕微鏡，アミノ酸分析装置，メタン測定装置）を提案した．

月惑星探査ロボットテストベッドのサブシステム（走行系，航法誘導系など）について報告するとともに，同ロボットを用いて，火山フィールドにおける実証試験の結果をまとめた．

A system for automated dust devil detection on Mars is developed based on images taken by the navigation camera aboard the Mars Exploration Rover, Spirit. This system enables us to pick up dust devils from the large number of images without checking each image visually and affords us a powerful tool for searching for dust devils and revealing their statistical properties for scientific studies. From an engineering point of view, the volume of dust devil image data to be downlinked to Earth for dust devil studies can also be drastically reduced by automated detection of dust devils because we can trim dust devil regions on the images onboard. A simple performance test using images taken on a particular sol shows that this system can detect almost all of the dust devils though also includes some false positives. Precision and recall are 0.85 and 0.81, respectively, which are comparable with or more effective than those in previous studies.

Soft soils cover on planetary surfaces, so the wheels for exploration rovers easily slip. Understanding an interaction between a wheel and soft soils is important for a traction control to improve rover traversability. The interaction between the wheel and soft soils has been studied in a field of terramechanics. Since the classical terramechanics-based wheel model considers only a static state of a wheel sinkage, the wheel model is not applicable to the wheel sinkage and slip. Thus the dynamic normal stress model and shear deformation model were proposed to deal with the problems of the wheel sinkage and the slip. The dynamic normal stress model was proposed by considering the wheel sinking velocity and a variation of soil state for solving the problem of wheel sinkage. And the shear deformation model was proposed by considering the shear characteristics for solving the problem of wheel slip. The shear deformation model was formulated by the shear test. The simulation verified the validity of the dynamic normal stress model and the shear deformation model in the transient state. In this paper, a single wheel experiments was performed to evaluate the shear deformation model by comparing the simulation results and the experimental results. Then the characteristics of the wheel sinkage and slip and the effectiveness of the model were confirmed.

（要旨） 本論文では，月縦孔降下中の探査プローブが撮像した複数の時系列画像を用いて，月縦孔側壁形状を3次元再構成することを目的とする．3次元再構成では，Simultaneous Localization and Mapping (SLAM) の技術を援用し，探査プローブの自由落下運動条件をSLAM の各処理過程に組み込むことによって，カメラの回転に頑強なSLAM へと改良した．提案SLAM のシミュレーション検証において，3次元再構成の精度向上を確認した．さらに，探査プローブ搭載カメラの性能の違いによる3次元再構成への影響の検証も実施し，本研究で得られた検証結果から，ミッションシナリオと探査プローブの運動要件，カメラ性能に関する議論を行った．

We are planning to explore the caverns through the skylight holes on the Moon and Mars. The holes and their associated subsurface caverns are among the most important future exploration targets. The importance of the lunar and Martian holes and their associated caverns is categorized from two aspects: (1) fresh materials are easily observed and sampled there, and (2) the subsurface caverns provide a safe, quiet environment. The expectation of lunar and Martian hole and cavern exploration is increasing in Japan. We name the project as UZUME (Unprecedented Zipangu (Japan) Underworld of the Moon Exploration) whose name is after a Japanese mythology. The ultimate purpose of the UZUME project is to investigate how to expand human activity and survival in space and on extraterrestrial bodies.

Unmanned construction machine working in dangerous environments such as construction sites and disaster areas has been developed. However, it is still necessary to improve its work efficiency especially during bulldozing and excavating soil. This research aims to develop a method for predicting soil deformation using machine learning. The feasibility of the proposed method is verified in a scenario where a simple bulldozing blade excavates soil. In the experiment, soil deformation at a front part of the blade is captured by multiple stereoscopic cameras. The camera provides depth data that are then converted to height field data. This dataset is fed to machine learning using Recurrent Neural Network (RNN) because soil deformation is continuous phenomena depending on time variation. The learned model for predicting soil deformation is confirmed in varied intrusion depth of the bulldozing blade.

Unmanned aerial vehicles (UAVs) have been deployed in various applications. The UAVs are limited in their flight duration owing to the payload, battery capacity, or wind effects. To address the issue, we propose an energy-efficient path planning method for UAVs that explicitly considers spatiotemporal wind speed. The proposed method exploits the wind-based energy consumption model as the cost function and also implements a variable-resolution grid in the graph search. The simulation study with 648 scenarios confirms the validity of the proposed method. Further, the statistical analysis of the simulation results summarizes the impact of wind and freight mass on UAV path plannings.

Traversability prediction enables safe and efficient autonomous rover operation on deformable planetary surfaces. Revealing spatial distribution from terrain geometry to rover slip behavior is key to assessing prospective traversability, but is hindered by insufficient in situ measurements on hazardous states due to conservative rover traverses. To achieve a more accurate prediction, this letter proposes a framework that actively learns latent traversability by exploring informative terrain under the constraints of stochastic rover slip. With a Gaussian process (GP) modeling the spatial distribution, we devise an iterative two-stage framework that gradually refines the model estimation, combining risk-aware informative path planning and GP updates by taking in situ measurements. The path planning stage employs our designed sampling-based algorithm to generate informative trajectories with fault-tolerant risk inference, while the GP is cautiously updated with traverse data to avoid rover immobilization. Chance constraint formulation is exploited in the framework to infer the stochastic reachability of informative regions. Through GP estimates reducing uncertainty, the algorithm incrementally reaches informative yet hazardous states along feasible trajectories. Simulation studies in rough terrain environments demonstrate that the proposed framework gathers informative traverse data while averting rover stuck situations to estimate the latent traversability model.

Constructing lunar infrastructures including landing pad and habitable base is essential for enhancing lunar exploration and utilization in the future. The lunar surface regolith, fine-grained loose soil, needs to be compacted and consolidated enough for securely constructing the infrastructures. In this paper, we present a traction control of compaction roller towed by a small mobile robot. The control method aims to minimize the relative drawbar pull between the small mobile robot and the compaction roller. We also manufactured two different shapes of compaction rollers and verified the relationship between the roller shape and control performance. The experimental results show that the proposed traction control effectively minimizes the relative drawbar pull for the spiral-shaped roller while it becomes unstable for the cylindrical-shaped roller because of the roller excites soil bulldozing or excavation during its rolling.

A lunar landing base is a key infrastructure for mass transportation from Earth to the Moon and for sustainable activities on the Moon in the future. Unmanned robots are expected to survey the soil stiffness of multiple points for securely constructing the landing infrastructure. There is a trade-off between the time or energy required for the survey, the number of survey points, and the completeness of the survey. This paper proposes a path planning for robotic soil investigation that determines the location and the number of additional survey points. The proposed method aims to obtain completeness of information related to the soil stiffness as well as to reduce the survey time. The simulation study reveals that the method achieves obtaining a certain level of completeness while it reduces the time required for the survey.

Tracked vehicles has been widely used in unstructured environments such as on uneven/deformable terrains. The mobility of the track unit consisting of interconnected metal plates is the key to enhance the safety of the vehicle. This study focuses on the track and soil interaction, particularly the wetted soil. The Distinct Element Method is applied for analyzing the interaction stress between the track and soil with varied moisture content and slip ratio. The analysis revealed that the significant stress is generated to the leading edge of the track unit and the stress distribution shifts toward the rear region of the track in moist soil.