Collect useful sentences in Papers (3)

Many researchers have proposed the use of micro air vehicles (MAVs) as a promising alternative to ground robot platforms for rescue tasks and a host of other applications.

MAVs are already being used in several military and civilian domains, including surveillance operations, weather observation, disaster relief coordination, and civil engineering inspections.

As a result, previous autonomous MAVs have been limited in their ability to operate in these areas.

We describe experimental assessments of first using onboard sensors to estimate the vehicle’s position and second using the same sensor data to build a map of the environment around the vehicle, a process generally called simultaneous localization and mapping (SLAM).

due to a combination of limited payloads for sensing and computation, coupled with the fast dynamics of air vehicles, the algorithms have generally been tested in simulation, on ground robots, or with sensor data first collected from a manually piloted MAV

leverages existing algorithms to balance the trade-offs imposed by GPS-denied flight.

In the ground robotics domain, many algorithms exist for accurate localization in large-scale environments; however, these algorithms are usually deployed on slow-moving robots that cannot handle even moderately rough terrain.

Limited payload reduces the computational power available onboard and eliminates popular sensors such as SICK laser scanners, large-aperture cameras, and high-fidelity inertial measurement units (IMUs)

Indirect position estimates Whereas MAVs will generally have an IMU, double-integrating acceleration measurements from lightweight MEMS IMUs results in prohibitively large position errors.

Fast dynamics MAVs have fast and unstable dynamics that result in a host of sensing, estimation, control, and planning implications for the vehicle.

Constant motion Unlike ground vehicles, a MAV cannot simply stop and perform more sensing or computation when its state estimates have large uncertainties. Instead, the vehicle is likely to be unable to estimate its position and velocity accurately, and as a result, it may pick up speed or oscillate, degrading the sensor measurements further.

In recent years, the development of autonomous flying robots has been an area of increasing research interest.

This research has produced a number of systems with a wide range of capabilities when operating in outdoor environments.

Although these are all challenging research areas in their own right, and pieces of the previous work (such as the modeling and control techniques) carry over to the development of vehicles operating without GPS

Whereas outdoor vehicles can usually rely on GPS, there are many situation in which relying on GPS would be unsafe, because the GPS signal can be lost due to multipath, satellites being occluded by buildings and foliage, or even intentional jamming.

In response to these concerns, a number of researchers have developed systems that rely on vision for control of the vehicle.

Although the systems developed by these researchers share many of the challenges faced by indoor or urban MAVs, they operate on vehicles that are orders of magnitude larger, with much greater sensing and computation payloads.

In addition, the outdoor environments tend to be much less cluttered, which gives greater leeway for errors in the state estimation and control

Although their MAVs are able to hover autonomously, they do not achieve any sort of autonomous goal-directed flight that would enable the systems to be controlled at a high level such that they could be built upon for more advanced autonomous applications

To enable tractable vision processing, this work has typically made strong (and brittle) assumptions about the environment.

Their applicability is therefore constrained to environments with specific features and does not allow for general navigation in GPS-denied environments.

However, none of these papers presented experimental results demonstrating the ability to stabilize all six degrees of freedom of the MAV using the onboard sensors, and all made use of prior maps, an assumption that is relaxed in this work.

The system presented in this paper was also discussed in Bachrach, He, and Roy (2009a, 2009b)

Here we present a more detailed analysis and evaluation of the algorithms, as well as significantly expanded experimental results in the field, including new results in large-scale indoor environments and outdoor flight in the urban canyon.


Reference:

Bachrach, A., Prentice, S., He, R., & Roy, N. (2011). RANGE–Robust autonomous navigation in GPS‐denied environments. Journal of Field Robotics, 28(5), 644-666.