AlphaPilot: autonomous drone racing

This paper presents a novel system for autonomous, vision-based drone racing combining learned data abstraction, nonlinear filtering, and time-optimal trajectory planning. The system has successfully been deployed at the first autonomous drone racing world championship: the 2019 AlphaPilot Challenge. Contrary to traditional drone racing systems, which only detect the next gate, our approach makes use of any visible gate and takes advantage of multiple, simultaneous gate detections to compensate for drift in the state estimate and build a global map of the gates. The global map and drift-compensated state estimate allow the drone to navigate through the race course even when the gates are not immediately visible and further enable to plan a near time-optimal path through the race course in real time based on approximate drone dynamics. The proposed system has been demonstrated to successfully guide the drone through tight race courses reaching speeds up to $${8}\,{\hbox {m}/\hbox {s}}$$ 8 m / s and ranked second at the 2019 AlphaPilot Challenge.

State estimation of tire-road friction and suspension system coupling dynamic in braking process and change detection of road adhesive ability

During vehicle braking, when vehicles move on the road with unknown road roughness elevation and unknown tire/road friction coefficient, fewer sensors shall be used for vehicle braking closed-loop control and braking distance prediction to obtain the dynamic states of the vehicle suspension and tire systems. In this paper, a vehicle dynamic model is established in Carsim software. Modify lump LuGre friction model and road roughness elevation model of four tires are proposed based on matlab. When vehicles brake on the road with time-varying split-μ, a braking control algorithm established in this paper. The road roughness elevation and the braking force of each tire are supplied to the vehicle dynamic model in Carsim. A state estimate algorithm of suspension system is proposed. The scheme for minimum sensor of this estimator is determined. A state estimate algorithm of the tire/road friction using only tire angular velocity information is proposed. When vehicles brake on the road with different levels of roughness, the influence of the number of installation groups of the sensors, the tire vertical stiffness deviations, and the measurement noise on the estimation error of the estimator is analyzed. When the vehicle is driving on the road with unknown adhesive ability, based on the estimator of tire/road friction using only tire angular velocity information, the tire/road friction internal state, the changes of road adhesive ability, and the vehicle velocity are estimated well.

Surface-constrained continuous-time extended Kalman filter: optimal estimation of a state of a constrained dynamic system-ball-rover rolling on a known surface

The Surface-constrained Continuous-time Extended Kalman Filter (SCEKF), derived in thesis, contains a novel approach for handling surface or equality constraints, in which the surface-constrained CEKF is the projection of the unconstrained CEKF onto the set of state estimate rates that satisfy the constraints. The filter is used for optimal estimation of a state of a ball rolling on a known surface with uneven elevation. The state consists of surface contact point and geometrical center positions, attitude and angular velocity of the ball. The dynamics of the ball is affected by "unknown" to the filter disturbances, due to off-center point mass and variable wind. Thesis includes derivations of the SCEKF and the constraint dynamics of a rolling ball. The numerical computation results show that the surface-constrained filter can produce an accurate state estimate of the rolling ball and demonstrate that the estimate is significantly better than that produced by unconstrained filter.

Surface-constrained continuous-time extended Kalman filter: optimal estimation of a state of a constrained dynamic system-ball-rover rolling on a known surface

The Surface-constrained Continuous-time Extended Kalman Filter (SCEKF), derived in thesis, contains a novel approach for handling surface or equality constraints, in which the surface-constrained CEKF is the projection of the unconstrained CEKF onto the set of state estimate rates that satisfy the constraints. The filter is used for optimal estimation of a state of a ball rolling on a known surface with uneven elevation. The state consists of surface contact point and geometrical center positions, attitude and angular velocity of the ball. The dynamics of the ball is affected by "unknown" to the filter disturbances, due to off-center point mass and variable wind. Thesis includes derivations of the SCEKF and the constraint dynamics of a rolling ball. The numerical computation results show that the surface-constrained filter can produce an accurate state estimate of the rolling ball and demonstrate that the estimate is significantly better than that produced by unconstrained filter.

Mechanisms linking the Labrador Sea with subtropical Atlantic overturning

<p>We analyze the causal chain linking sea surface buoyancy anomalies in the Labrador Sea and variability in the subtropical Atlantic meridional overturning circulation (AMOC) in the ECCO ocean state estimate on inter-annual timescales. Our study highlights the importance of Lower North Atlantic Deep Water (LNADW) for the north-south connectivity in the Atlantic Ocean. We identify important mechanisms that allow the Labrador Sea to impact the southward transport of LNADW. We show that NAC plays an essential role in the export of buoyancy anomalies from the Labrador Sea – and it furthermore exerts a positive feedback that amplifies these upper ocean anomalies in the eastern subpolar gyre – before they reach the denser water masses along the lower limb of the AMOC. Our results also highlight the contribution of the western Labrador Sea for the surface uptake of tracers that penetrate the LNADW near Denmark Strait, which has implications for the redistribution of ocean heat anomalies.</p>