CASCADE Case Study: Scan a Bridge

• Single-operator aerial inspection of the Clifton Suspension Bridge

• Thermal data, workflow and business case

Operational Challenges: Safety and Agility

• GGG

245ft

76ft

331ft

Operational Challenges: Safety and Agility

• Highly complex environment

• Risk mitigations• Exploit road closures

for events and planned maintenance

• Railway rarely used• Bridge Trust and

Bristol City Council supportive

• Tether?

Paths

Railway

Busy roads

Complexity

• Mitigation: use off-the-shelf small (<7kg) multirotor

• DJI * or AscTec Falcon8• Limited on-board complexity

• Congested, complex environment

• Mostly VLOS, some EVLOS

• Category A or B operation• Depends how you judge the

complexity

Research Challenges

• Major goal: single person (E)VLOS operation in complex* environment• “Crew – usually minimum of two. One pilot in command, and one observer/camera

operator. Other missions may require more crew; spotters, engineers, aides and so on.” – Network Rail

• * Means we will not sterilise the area… but maybe it’s quiet and partly controllable

• Approach: on-ground autonomy to mitigate environment complexity• Mission design for safety

• Includes Remote Pilot location, spotting, contingencies, and robust mission execution• Autonomous execution monitoring by Remote Pilot Station

• Concept: autonomous RPA Observer for EVLOS• Simple – low workload – Remote Pilot interface

• Pause / Go / Abort

System Platform Payload

ControllerAutonomous

AssistantMobile

Assistant

Tracking

Area Monitoring

Tracking

Mission Planner

H

Mission Design

• Inputs• CAD model of target structure• Routes of uncontrolled users through

region of interest• Platform and sensor parameters

• Output: robust sequence of segments• E.g. a behaviour tree• Each has safety preconditions

• Battery > X%; Zone Z clear; UAV at W; Remote Pilot at location Y

• System integrity confirmed• Whole mission proven safe

• Contingencies pre-planned

This Photo by Unknown Author is licensed under CC BY-NC-SA

Behaviour

DMS

Blackboard Tasks

States Sensing Dashboard

getPose getTwist getOtherCars approach_warning clear_range Passed car check LR pedestrian check RL pedestrian check RC pedestrian check OC cyclist check Button1_2BB Button2_2BB Button3_2BB Button4_2BB

Laps

Lap 1 Lap 2 Lap 3 Lap 4 Lap 5 Lap 6

drive L1 P1

Drive North -> Steps

Right - Steps - Empty drive L2 P1

Always True Drive Steps -> Steps

Right - Steps - Ped drive L3 P1

road crossing ped Drive Steps -> Steps*

Right - Steps - cyclist drive L4 P1 Left - MOD drive L4 P2

OC car Drive Steps -> MOD Always True* Drive MOD -> HP

Right - HP drive L5 P2 Right - MOD

Always True** Overtake Drive HP -> MOD

Overtake? Overtake?*

Request overtake Overtake right

Overtake Button? clear range check OC car* Passed?

Passed Obstacle CL Right

passed car CL Left

on overtake straight? running_is_failure

passed car*

Request overtake* Overtake right*

Overtake Button?* clear range check* OC car** Passed?*

Passed Obstacle* CL Right*

passed car** CL Left*

on overtake straight?* running_is_failure*

passed car***

Always True***

drive L6 P1 Right - Steps drive L6 P2 Right - HP* drive L6 P3 Zebra Crossing drive L6 P4

Drive Mod -> Steps Always True**** Drive steps -> HP Always True***** Overtake* Drive HP -> crossing

Overtake?**

Request overtake** Passed?**

Overtake Button?** clear range check** Passed Obstacle** CL Right**

passed car**** CL Left**

on overtake straight?** running_is_failure**

passed car*****

crossing ped RL Drive crossing -> North

Nuggets

MAIN STREAM• Mission modelling and design

methods

• Statistical basis for acceptance trials criteria

• SUA platform integration and safety case

• Autonomous Remote Pilot Station

EXTRA SCOPE (PhDs? Creep?)• Autonomous RPA Observer

• Flight design and operations with safety tether

• Ballistic characteristics of standard platforms

• Smartphone-based black box• “Internet of Drones”

Demonstration Milestones

• Following nuclear industry’s“lab / white / red” concept

• Thorough statistical analysis• Acceptance criterion = X hours of

problem-free operation• “Problem” here means an

unhandled circumstance• Manual intervention?

• Quick & dirty demo in Aug 2018

• Lab• Simulation HWIL sim HW at flying site

• Simulated disturbances, failures and passers-by

• White• Real site, but with closures, and

trained participants interacting• Include simulated failures

• Red• Real site, real day

Collaborations

Internal

• Cranfield: formal mission spec

• Manchester: mission & safety

• Southampton: visual tracking

• Imperial: drones & tethers

External

• UK Collaboratorium for Research on Infrastructure and Cities (UKCRIC):

• Value added and payload

• Clifton Suspension Bridge Trust• A bridge!

• Assuring Autonomy (?)

Next Steps

• “Quick and dirty” concept experiment

• August 2018

• Refinement of scope• Payload: UKCRIC• Tech: CASCADE

• Possible aligned PhDs• Sept 2018/19 start

• Target: August 2020

• What’s not in scope?

• Turbulence & poor weather

• Precision proximity control

• Multi-platform operations

• Autonomous flight control

• Dynamic mission re-planning

• 5G connectivity