Relative Navigation: A Keyframe-Based Approach for Observable GPS-Degraded Navigation

Wheeler, David O.; Koch, Daniel P.; Jackson, James S.; McLain, Timothy W.; Beard, Randal W.

doi:10.1109/mcs.2018.2830079

Cited by 26 publications

(30 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…6 illustrates through Monte Carlo simulations the empirical distribution of the estimated pose of the egovehicle when the communicated pose (in black) and the relative pose have Gaussian distributions. One can see that the results are actually following banana-shaped distributions [20], [19] which are badly approximated by Gaussian distributions when they are large. The first formulation (in blue) gives a distribution less Gaussian than the second one (in red) because of the dependency of the relative orientation error, see Eq.…”

Section: B Propagation Accuracy and Consistency 1) Lidar Detectionsmentioning

confidence: 98%

Pose and covariance matrix propagation issues in cooperative localization with LiDAR perception

Héry

2019

2019 IEEE Intelligent Vehicles Symposium (IV)

View full text Add to dashboard Cite

This work describes a cooperative pose estimation solution where several vehicles can perceive each other and share a geometrical model of their shape via wireless communication. We describe two formulations of the cooperation. In one case, a vehicle estimates its global pose from the one of a neighbor vehicle by localizing it in its body frame. In the other case, a vehicle uses its own pose and its perception to help localizing another one. An iterative minimization approach is used to compute the relative pose between the two vehicles by using a LiDAR-based perception method and a shared polygonal geometric model of the vehicles. This study shows how to obtain an observation of the pose of one vehicle given the perception and the pose communicated by another one without any filtering to properly characterize the cooperative problem independently of any other sensor. Accuracy and consistency of the proposed approaches are evaluated on real data from on-road experiments. It is shown that this kind of strategy for cooperative pose estimation can be accurate. We also analyze the advantages and drawbacks of the two approaches on a simple case study.

show abstract

Section: B Propagation Accuracy and Consistency 1) Lidar Detectionsmentioning

confidence: 98%

Pose and covariance matrix propagation issues in cooperative localization with LiDAR perception

Héry

2019

2019 IEEE Intelligent Vehicles Symposium (IV)

View full text Add to dashboard Cite

show abstract

“…As shown in Ref. 7, methods that directly estimate the global state are susceptible to inconsistency and state jumps during prolonged GPS dropout. The value of a relative parameterization is welldocumented in the full-SLAM literature, 5,[25][26][27] but has not been fully applied to MAV navigation.…”

Section: Related Workmentioning

confidence: 99%

“…5,6 Significant reliability issues arise when working with respect to a globallyreferenced state during prolonged GPS dropout and heading uncertainty. 7 Despite these issues, many current GPS-denied MAV navigation approaches continue to estimate and control with respect to a single, inertial reference frame: either the GPS origin or the MAV's initial pose. This formulation is convenient; however, there are several underlying issues that commonly arise in GPS-degraded environments when estimation and control are carried out with respect to a global reference frame:…”

Section: Introductionmentioning

confidence: 99%

“…• During prolonged GPS dropout, the unobservable global states become inconsistent, 5,7 resulting in a poor understanding of the uncertainty of the vehicle's global pose. Inconsistency reduces estimator optimality, 6 can cause the estimator to gate valid GPS measurements if GPS is eventually reacquired, and can negatively impact applications such as geofencing that require a good understanding of the global uncertainty.…”

Section: Introductionmentioning

confidence: 99%

“…The relative state estimator ensures that the state is observable and the uncertainty remains bounded, consistent, and normally-distributed. 7 By removing the globalstate estimation from the front end, RN also ensures that large or delayed global-state updates, which come from incorporating loop-closure constraints or eventual global measurements, do not impact the flight-critical control and estimation feedback. Rather, the global state is estimated independently using a pose-graph back end where the non-Gaussian uncertainties can be better represented and robust optimization methods can identify and reject gross GPS outliers and false-positive loop closures.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Relative navigation of autonomous GPS-degraded micro air vehicles

et al. 2020

Self Cite

View full text Add to dashboard Cite

Many current approaches for navigation of micro air vehicles (MAVs) in GPS-degraded environments use a globally-referenced state for estimation and control, even though this state is not observable when GPS is unavailable. By working with respect to a local reference frame, the relative navigation (RN) framework presented in this paper ensures that the state maintains observability and that the uncertainty remains bounded, consistent, and normally-distributed. RN further insulates flight-critical estimation and control processes from the large global updates common in GPSdegraded MAV flight. This paper provides a thorough description of the details needed to successfully implement the RN framework on a MAV. The practicality of RN is demonstrated in several long flight tests in unknown, GPS-denied and GPS-degraded environments. The relative front end is shown to produce low-drift estimates and smooth, stable control while leveraging off-the-shelf algorithms. The system runs in real time with onboard processing, fuses a variety of vision sensors, works indoors and outdoors, and does not require special tuning for particular sensors or environments. RN is also shown to produce globally-consistent, metric, and localized maps by incorporating loop closures and intermittent GPS measurements. This map is used to demonstrate autonomous completion of mission objectives. By subtly restructuring the estimation framework, RN promotes a paradigm shift that avoids many issues inherent in GPS-degraded navigation.

show abstract