#178 Aerial Contact Sensing for Improved Inspection of Transportation Infrastructure

Principal Investigator
Sebastian Scherer
Start Date
July 1, 2018
End Date
June 30, 2019
Research Type
Grant Type
Grant Program
FAST Act - Mobility National (2016 - 2022)
Grant Cycle
2018 Mobility21 UTC


The same text as below is uploaded as a PDF document with figures and nicer formatting. Also attached as a document is a letter of support from Near Earth Autonomy.

Contact measurements are one of the key tasks performed during inspection of transportation infrastructure. Currently these tasks require an inspector to manually climb onto the structure and record the measurements. We propose to improve the efficiency of these measurements by developing an arm capable of contact attached to a flying robot. We expect the novel contributions to be a lightweight compliant arm mechanism and novel control and planning algorithms for a flying robot to enable contact inspection. The developed system will be deployed and tested on several bridges in Pittsburgh.    
We propose to increase the utility of aerial infrastructure inspection of bridges beyond the current state of the art where imagery typically is manually collected from structures at a distance of many meters with research to enable contact sensing. These contact sensing technologies will enable safer, more cost-effective sUAS inspection of transportation infrastructure than is possible by existing human manual inspection. 

Bridge inspection today is performed by workers or manually controlled robots and drones to find abnormal indications such as gaps, cracks, buckling, and corrosion. These efforts often entail lengthy bridge shutdown and construction of scaffolding for inspectors. Visually apparent defects as well as readings taken from contact sensors are noted on paper. We believe our automated contact sensing will radically increase up-time while improving personnel safety. 

Our proposed approach extends key innovations from an NSF National Robotics Initiative project at Carnegie Mellon and Northeastern University focused on bridge inspection using aerial robots. This project was highlighted during the Frontiers Conference led by President Obama in October 2016. 
Our innovation proposed for the Mobility21 Initiative is based on researching and examining a manipulator and control algorithms that can be deployed from a UAS to make measurements that are in contact with, the surfaces being inspected.

Challenges and Objective

This proposal addresses the key challenge of how to safely and effectively perform contact sensing from a small aircraft

There are established contact sensors available commercially however it is a challenge to achieve effective readings using typically hand-held sensors on an aircraft. The sensors must reach out away from the aircraft to contact structures and must maintain contact for some amount of time. Simple implantations can cause the vehicle to become unstable during contact and the vehicle needs to be actively controlled to provide good contact.

Objective: Facilitate collection of coating thickness, structural integrity and/or other data from a sUAS using contact sensors.

Questions to be answered: Which contact sensor types are feasible and of most value to accommodate? Can ultrasonic thickness gauges, Schmidt hammers, and/or other common sensors be accommodated? How are contact sensors best held and positioned by an sUAS? How can we ensure safe flight and maintain an ease of use?

Technical Approach

We will design a lightweight, smart mechanism to safely position sensors against structures. These mechanisms will account for the motions inherent within the aircraft position control envelope. We intend to use passive compliance to account for the motions while maintaining sensor contact. Contact sensors will allow the aircraft controls to monitor contact status and know when to record sensor data. Additional dedicated range sensors can help keep the aircraft safe when it is close to structures but not yet in contact. In previous work, we have explored various mechanism for contact  and will use that expertise to apply it to this project.
When GPS is either intermittent or unavailable we use lidar and vision-based navigation methods. The sensor data allows the onboard system to model the environment and estimate aircraft state, i.e. position and orientation, relative to the nearby structures. 

One of the key technological hurdles to safe flight indoors is reliable position and velocity estimation. Since the small inertial measurement units that can be carried onboard cannot be solely used to determine the position and velocity it is necessary to instead rely on an external reference with respect to the structure that will be inspected. However, since the sensors onboard of the aircraft have limited range and field of view it is not always possible to estimate the position at any location with respect to the structure. While vision-based methods to estimate the position and orientation are very popular they are of limited utility inside structures, because the lighting is often poor and textures are often repetitive. Instead, we utilize a spinning lidar scanner to localize with respect to a self-created map of the structure.
The table below defines the R&D plan in terms of tasks and associated deliverables, and the corresponding schedule. Each development will include design, implementation, testing, and iteration. We have flying robots, actuators, and materials already available to start on the project and don’t require equipment funds.
Key milestones are: Month 6 – First in-lab demonstration of contact sensing. Month 12 – flight demonstration in relevant environment. 

Task                                                                                                            	Deliverable(s)
1	Develop and mount manipulator                                                     	 Development report
2	Develop flight maneuvers and control to enable manipulation          	Development report
3	Evaluate approach in experimentation                                               	Demonstration video and verification report

Project Quarter:	                  Q1	Q2	Q3	Q4
Task 1: Develop manipulator	       x.       x 			
Task 2: Develop control				         x
Task 3: Evaluate Approach				         x
Deployment Plan
An initial deployment of the system will be at the Charles Anderson Bridge in Schenley Park to measure the thickness of steel. We have previously arranged with the City of Pittsburgh and PennDot to test at this bridge. Once we have started initial deployments on these bridges we will test the system on larger structures together with PennDOT and gather their feedback.    
Expected Accomplishments and Metrics
•	A key new capability to get contact measurements without requiring special infrastructure such as cranes or scaffolding to reach these locations.
•	A new mechanism design for contact inspection and new algorithms to enable control.
•	A deployment and evaluation of the system on a real structure to gather measurements.

Individuals Involved

Email Name Affiliation Role Position
basti@andrew.cmu.edu Scherer, Sebastian Robotics Institute PI Faculty - Untenured, Tenure Track
weikunz@andrew.cmu.edu Zhen, Weikun MechE Other Student - PhD


Amount of UTC Funds Awarded
Total Project Budget (from all funding sources)


Type Name Uploaded
Data Management Plan datamanagementplan.pdf Jan. 12, 2018, 2:18 p.m.
Progress Report 178_Progress_Report_2018-09-30 Sept. 28, 2018, 12:39 p.m.
Progress Report 178_Progress_Report_2019-03-30 March 4, 2019, 9:34 a.m.
Publication Mobility21_Report_Scherer.pdf Aug. 1, 2019, 3:55 a.m.
Progress Report 178_Progress_Report_2019-06-30 July 31, 2019, 3:42 p.m.
Final Report Mobility21_Report_RRW6svO.pdf Aug. 1, 2019, 4:27 a.m.
Progress Report 178_Progress_Report_2019-09-30 Sept. 30, 2019, 11:07 a.m.
Publication Autonomous exploration for infrastructure modeling with a micro aerial vehicle April 19, 2021, 7:54 a.m.

Match Sources

No match sources!


Name Type
Near Earth Autonomy Deployment Partner Deployment Partner