Unsupervised classification of slip events for planetary exploration rovers

Bouguelia, Mohamed-Rafik; González, Roberto J.; Iagnemma, Karl; Byttner, Stefan

doi:10.1016/j.jterra.2017.09.001

Cited by 32 publications

(18 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another interesting discussion about this topic comprises the selection of supervised machine learning versus unsupervised machine learning approaches. In the context of planetary exploration robotic missions, unsupervised learning methods are preferred over supervised ones as identifying correct responses, required for training supervised methods and creating a model of the process, can be risky (e.g., high slip events) and time consuming (manual labeling from the Earth‐supporting team) …”

Section: Conclusion and Future Challengesmentioning

confidence: 99%

Slippage estimation and compensation for planetary exploration rovers. State of the art and future challenges

González

Iagnemma

2017

Journal of Field Robotics

Self Cite

View full text Add to dashboard Cite

Future lunar/planetary exploration missions will demand mobile robots with the capability of reaching more challenging science targets and driving farther per day than the current Mars rovers. Among other improvements, reliable slippage estimation and compensation strategies will play a key role in enabling a safer and more efficient navigation. This paper reviews and discusses this body of research in the context of planetary exploration rovers. Previously published state‐of‐the‐art methods that have been validated through field testing are included as exemplary results. Limitations of the current techniques and recommendations for future developments and planetary missions close the survey.

show abstract

Section: Conclusion and Future Challengesmentioning

confidence: 99%

Slippage estimation and compensation for planetary exploration rovers. State of the art and future challenges

González

Iagnemma

2017

Journal of Field Robotics

Self Cite

View full text Add to dashboard Cite

show abstract

“…Section 4 explains the data set (data collection) and the set of features considered for training the machine learning algorithms. Section 5 provides experimental results showing a comparison among the regression algorithms presented in this work and other algorithms previously published by the authors and based on machine learning classification (Bouguelia et al, 2017;Gonzalez et al, 2018). Finally, conclusions and future work are drawn in Section 6.…”

Section: Introductionmentioning

confidence: 99%

Slippage prediction for off-road mobile robots via machine learning regression and proprioceptive sensing

González

Fiacchini

Iagnemma

2018

Robotics and Autonomous Systems

Self Cite

View full text Add to dashboard Cite

This paper presents a new approach for predicting slippage associated with individual wheels in off-road mobile robots. More specifically, machine learning regression algorithms are trained considering proprioceptive sensing. This contribution is validated by using the MIT single-wheel testbed equipped with an MSL spare wheel. The combination of IMU-related and torque-related features outperforms the torque-related features only. Gaussian process regression results in a proper trade-off between accuracy and computation time. Another advantage of this algorithm is that it returns the variance associated with each prediction, which might be used for future route planning and control tasks. The paper also provides a comparison between machine learning regression and classification algorithms.

show abstract

“…The problem of detecting degraded mobility, in particular in the form of high rover slip, is an important aspect of avoiding immobililization. One line of work in this area detects high slip events through classification of proprioceptive data (e.g., wheel torque and inertial measurement unit [IMU], data) without the need for absolute positioning that can require complex visual processing (e.g., VO; Bouguelia, Gonzalez, Iagnemma, & Byttner, ; Gonzalez, Apostolopoulos, & Iagnemma, ; Iagnemma & Ward, ). Currently, NASA’s Curiosity rover uses VO to compute slip and employs a static slip threshold to automatically stop a traverse if a “fast” slip threshold is exceeded even once or if a “slow” slip threshold is exceeded over consecutive measurements (Arvidson et al, ).…”

Section: Introductionmentioning

confidence: 99%

Data‐driven mobility risk prediction for planetary rovers

Skonieczny

Shukla

Faragalli

et al. 2018

Journal of Field Robotics

View full text Add to dashboard Cite

Mobility assessment and prediction continues to be an important and active area of research for planetary rovers, with the need illustrated by multiple examples of high slip events experienced by rovers on Mars. Despite slip versus slope being one of the strongest and most broadly used relationships in mobility prediction, this relationship is nonetheless far from precisely predictable. Although the literature has made significant advances in the predictability of average mobility, the other key related aspect of the problem is the risk caused by edge cases. A key contribution of this study is a metric for explicitly assessing mobility risk based on data-driven nonparametric slip versus slope relationships. The data-driven approach is meant to address limitations of past model-based approaches. The metric is informed by past work in terramechanics relating drawbar pull (i.e., net traction) to slip: High slip fraction (HSF), defined as the proportion of slip data points above 20%. Another contribution is a low complexity mobility prediction framework, the autonomous soil assessment system. Field tests demonstrate that, for sand and gravel, rover trafficability becomes nonlinear and highly variable above the 20% slip threshold.HSF is shown to be a useful metric for categorizing rover-terrain interactions into low, medium, or high risk, correctly and consistently. Furthermore, the metric is shown to be useful for early detection of potentially hazardous changes in roverterrain conditions. The combination of HSF with an appropriately sized queue structure for modeling slip versus slope enables an appropriate balance between responsiveness and stability. K E Y W O R D S mobility prediction, planetary robotics, rover slip

show abstract

Unsupervised classification of slip events for planetary exploration rovers

Cited by 32 publications

References 16 publications

Slippage estimation and compensation for planetary exploration rovers. State of the art and future challenges

Slippage estimation and compensation for planetary exploration rovers. State of the art and future challenges

Slippage prediction for off-road mobile robots via machine learning regression and proprioceptive sensing

Data‐driven mobility risk prediction for planetary rovers

Contact Info

Product

Resources

About