Robert C. Leishman

Quadrotors and Accelerometers: State Estimation with an Improved Dynamic Model

Results are presented that quantify how velocity and attitude estimates can benefit from an improvement to the traditional quadrotor dynamic model. The improved model allows accelerometer measurements, which are routinely available at high rates, to reduce an estimators dependence on complex exteroceptive measurements, such as those obtained by an onboard camera or laser range finder.

Relative navigation approach for vision-based aerial GPS-denied navigation

GPS-denied aerial flight is a challenging research problem and requires knowledge of complex elements from several distinct disciplines. Additionally, aerial vehicles can present challenging constraints such as stringent payload limits and fast vehicle dynamics. In this paper we propose a new architecture to simplify some of the challenges that constrain GPS-denied aerial flight. At the core, the approach combines visual graph-SLAM with a multiplicative extended Kalman filter. More importantly, for the front end we depart from the common practice of estimating global states and instead keep the position and yaw states of the MEKF relative to the current node in the map. This relative navigation approach provides simple application of sensor measurement updates, intuitive definition of map edges and covariances, and the flexibility of using a globally consistent map when desired. We verify the approach with hardware flight-test results.

Analysis of an Improved IMU-Based Observer for Multirotor Helicopters

Multirotor helicopters are increasingly popular platforms in the robotics community. Making them fully autonomous requires accurate state estimation. We review an improved dynamic model for multirotor helicopters and analyze the observability properties of an estimator based on this model. The model allows better use of IMU data to facilitate accurate state estimates even when updates from a sensor measuring position become less frequent and less accurate. We demonstrate that the position update rate can be cut in half versus typical approaches while maintaining the same accuracy. We also find that velocity estimates are at least twice as accurate independent of the position update rate.

Relative navigation and control of a hexacopter

This paper discusses the progress made on developing a multi-rotor helicopter equipped with a vision-based ability to navigate through an a priori unknown, GPS-denied environment. We highlight the backbone of our system, the relative estimation and control. We depart from the common practice of using a globally referenced map, preferring instead to keep the position and yaw states in the EKF relative to the current map node. This relative navigation approach allows simple application of sensor updates, natural characterization of the transformation between map nodes, and the potential to generate a globally consistent map when desired. The EKF fuses view matching data from a Microsoft Kinect with more frequent IMU data to provide state estimates at rates high enough to control the vehicles fast dynamics. Although an EKF is used, a nodes and edges graph represents the map. Hardware results showing the quality of the estimates and flights with estimates in the loop are provided.

Differential flatness based control of a rotorcraft for aggressive maneuvers

We propose a new method to control a multi-rotor aerial vehicle. We show that the system dynamics are differentially flat. We utilize the differential flatness of the system to provide a feed forward input. The system model derived allows for arbitrary changes in yaw and is not limited to small roll and pitch angles. We demonstrate in hardware the ability to follow a highly maneuverable path while tracking a time-varying heading command.

Multiplicative Extended Kalman Filter for Relative Rotorcraft Navigation

This paper details the fundamentals of a new approach to navigation for aerial vehicles in confined indoor environments without access to global-position measurements. The approach departs from the common practice of navigating within a globally referenced map, and it instead keeps the position and yaw states relative to the current node in the map. The approach combines elements of graph-based simultaneous localization and mapping with a multiplicative extended Kalman filter. The filter provides accurate state estimates at a fast rate and provides the information necessary for a simultaneous localization and mapping algorithm to maintain a pose graph. Specific details for the relative multiplicative extended Kalman filter are provided. The relative estimation approach is validated with hardware flight tests, and results are compared to motion capture ground truth data. In addition, flight-test results using estimates in the control loop are provided.

Robust Motion Estimation with RGB-D Cameras

Estimating vehicle motion using vision sensors in real time has been greatly explored in the past few years due to speed improvements and advances in computer hardware. Six degree of freedom motion estimation using vision information is desirable due to a vision sensors low cost, low power requirements and light weight and for the quality of the solutions that can be obtained using few assumptions about the environment. However, cameras have the downside of not providing good estimates when visual features are sparse or not available. Also, there are problems with changes in lighting and when light is low or unavailable. Laser scanners have been shown to be robust in these situations. We view an RGB-D sensor as providing three complimentary modalities that are useful for providing motion estimation solutions: a monocular camera, a 3D point cloud and the combination providing RGB-D information. Obviously motion estimates produced using the combined sensor information are best. However, there are times when information from both sensors is not available. The monocular camera remains useful when depth information is absent or insufficient, like in a large room, down a long hallway or outdoors. The 3D point cloud may still be available when there is insufficient light to utilize the RGB image. The approach described in this work seeks to take advantage of all three of these sensor modalities to provide a more robust motion estimation solution.

Utilizing an improved rotorcraft dynamic model in state estimation

Multirotor aircraft have become a popular platform for indoor flight. To navigate these vehicles indoors through an unknown environment requires the use of a SLAM algorithm, which can be processing intensive. However, their size, weight, and power capacity limit the processing capabilities available onboard. In this paper, we describe an approach to state estimation that helps to alleviate this problem. By using an improved dynamic model we show how to more accurately estimate the aircraft states than can be done with the traditional approach of integrating IMU measurements. The estimation is done with relatively infrequent corrections from accelerometers (40Hz) and even less frequent updates from a vision-based SLAM algorithm (2–5 Hz). This benefit of requiring less frequent updates from processing intensive sources comes without significant increase in the estimators complexity.