Improved Autonomy for Unmanned Aerial Vehicles in Unstructured Environments

预计全球无人机（UAV）支出将从2015年的40亿美元大幅增长到2025年的140亿美元，十年内总计930亿美元（来源：Teal Group）。军事研究预计将在这十年中增加300亿美元的支出。这种显著的增长是因为无人机扩展和补充了人类的能力，使其能够更快，更准确，更安全地完成困难的工作，并减少预算。民用应用包括在人类无法生存的地区进行灾难恢复，或检查基础设施，如覆盖大片无法进入的土地的输电线路。无人机是军事机器人革命的一部分，它们通常用于情报，监视和侦察（ISR）。它们是全球军事力量未来的关键，特别是加拿大空军的未来，正如加拿大空军空战中心在战略规划文件中提到的那样，如“2019年塑造加拿大空军的预测力量趋势”。 拟议研究的目的是提高无人机在非结构化环境中的自主性;这项工作建立在申请人2004年开始的过去成功的无人机研究计划的基础上。我们将自主性广义地定义为无人机在一定程度上独立操作的人机团队所实现的能力。对已知环境的自主控制（例如，具有信标的映射仓库）不能应用于非结构化环境。在那里，机器人必须感知它们的环境，以便为使命目标做出智能决策。首先，我们研究了在GPS退化的环境中（例如，由于对抗性干扰）。同样重要的情况是相对于具有未知GPS坐标的物体的导航（例如，抓取有效载荷）。虽然基于视觉的导航已被用于视觉伺服，但它可能是不必要的计算密集型的，因为它依赖于对特征的连续跟踪和地图的维护。因此，第二个目标是研究用于特定自主运动控制任务的有效视觉伺服，包括在移动的船舶甲板上着陆。第三，对无人机控制中的时滞进行了分析和补偿。视觉伺服和导航或远程遥控操作中所需的图像处理引入了显著的延迟，这限制了性能。第四，研究了多车协调控制的一般原理及其在吊具运输中的应用。 该研究为非线性控制、视觉伺服、视觉导航和机器人等领域提供了新的、广泛适用的理论，具有重要意义。这一理论在具有实际重要性的具体应用中得到了证明。对加拿大的好处包括在大型和不断增长的无人驾驶系统领域的创新，这些系统正在改变军事和民用应用。

Global spending on unmanned aerial vehicles (UAV) is predicted to grow dramatically from USD 4 billion in 2015 to USD 14 billion by 2025, which is a total of USD 93 billion in a decade (source: Teal Group). Military research is expected to add USD 30 Billion in spending over this decade. This remarkable growth is because UAVs extend and complement human capability to do difficult jobs faster, more accurately, safer, and at a reduced budget. Civilian applications include disaster recovery in regions where humans cannot survive, or inspection of infrastructure such as electrical transmission lines covering vast expanses of inaccessible land. UAVs are part of a robotics revolution in the military where they are often used for intelligence, surveillance, and reconnaissance (ISR). They are key to the future of military forces worldwide, and in particular to that of Canada's Air Force as mentioned in strategic planning documents such as “Projecting Power Trends Shaping Canada's Air Force in the Year 2019” by the Canadian Forces Air Warfare Center. The aim of the proposed research is to improve UAV autonomy in unstructured environments; the work builds on the applicant's past successful UAV research program begun in 2004. We define autonomy broadly as capabilities achieved by a man-machine team where the UAV operates with some level of independence. Autonomous control for known environments (e.g., a mapped warehouse with beacons) cannot be applied to unstructured environments. There, robots must perceive their environment to make intelligent decisions towards a mission goal. First, we investigate new methods for vision-based navigation in an environment where GPS is degraded (e.g., due to adversarial jamming). An equally important case is navigation relative to an object with unknown GPS coordinate (e.g., grasping a payload). Although vision-based navigation has been used for visual servoing, it can be unnecessarily computationally intensive as it relies on continuous tracking of features and the maintenance of a map. Therefore, a second objective researches efficient visual servoing for specific autonomous motion control tasks including landing on a moving ship deck. Thirdly, we analyze and compensate for time delay in UAV control. Image processing required in visual servoing and navigation or long-range remote teleoperation introduces significant latency which limits performance. Fourthly, we investigate multi-vehicle coordinated control in general and its application to slung load transportation. The proposed research is important as it develops novel broadly applicable theory in the field of nonlinear control, visual servoing, visual navigation, and robotics. This theory is demonstrated on specific applications of practical importance. The benefits to Canada include innovation in the large and growing field of unmanned systems which are transforming military and civil applications.