Joseph Robinson, Robert Watson, Sam Coe, Matt van Berlo, Josh ...
Undergraduate/Graduate Category: Degree Level: Undergraduate Abstract ID# 567
i93 Tunnel Inspection Robot
Joseph Robinson, Robert Watson, Sam Coe, Matt van Berlo, Josh Johnson Advisor: Professor Shafai Berham Abstract
State of the Art
Robot System
A mobile robo@c system has been built as a tool to improve the bridge inspec@on process done by the MassachuseDs Department of Transporta@on (MassDOT). Current methods for inspec@ng the plenums (ceiling space) are @me consuming, costly, and pose significant safety hazards to both the inspectors and the drivers. Our system will provide a means for inspectors to conduct these inspec@ons at the seat of their desks. On top of reducing cost, @me, and poten@al hazards, the mobile plaPorm enhances inspec@on reports with its soQware package offering automated & vision technologies.
• •
(b)
Inspec@ons are done while inspectors crawl in plenum place, and documen@ng notes and hand sketches are done this way for the en@re length [see Fig 3]. Lane closures are held for inspectors to enter and exit the plenum crawl space.
Head Light and Camera Flash
LiV Camera (Servo Controlled)
Background Star@ng back in 1991, the Big Dig remapped the 6-‐laned I-‐93 from above the city to below via a 3.5 mile long tunnel. Inten@on was to rid Boston of severe traffic concerns, which caused up to ten hours of traffic per day for city drivers. Projec@ons forecasted traffic reaching up to sixteen hours a day by 2010. In 2006, a 12t ceiling @le fell on a car and the drivers. AQer, several anchors embedded in the tunnel’s roof slab were discovered to be the cause. City-‐wide tunnel inspec@ons this problem as abundant, even in newly constructed areas [see Fig 1]. Fig 1 Showing the ceiling collapse (leQ) and simulated view of the cause (right).
MassDOT spends over $3M each year on plenum inspec@ons. Inspec@ons call for lane closures, reintroducing traffic conges@ons. Plus, working condi@ons have proven unsafe for both the inspectors and the drivers alike.
(a)
(b)
(c)
Fig 3 (a) A picture of the i93 plenum space inspected each year. (b) Entrance to Plenum crawl space. (c) Our envision of taking the inspector out of the tunnel and into his desk while performing inspec@ons.
LiV PlaXorm (3D Printed)
• Lane closures are held for inspectors to enter and exit the plenum crawl space. Base Camera w Bumper Protectors
Motor Controls
Results Fig 4 A close look at the prototyped Mobile System
Pause/Play Media
Camera 2 Control
• •
Mobile PlaPorm has been built for less than $1K [see Fig 4].
• •
Servos orientated parallel to ver@cal & horizontal results in 2D rota@onal control.
Launch Viewer
Camera 2 i/o
Set Flag/ Next Hanger
•
Record Video
Xbox controller provides an ease of use [see Fig 5].
Graphical User Interface with embedded media stream together with func@ons, features, and system informa@on [see Fig 7]. Google SketchUp model incorpora@ng accurate geometries and structural layout, along with the appropriate physics governing robot’s mo@on [see Fig 8 & 9].
Take Picture
Fig 6 Servos orthogonal to one another, providing rota@on in the ver@cal and horizontal. Base Control
Method
Base Control i/o Controller i/o
• • •
Provide MassDot a mobile plaPorm controllable from a PC with WiFi.
•
Build model for offline test simula@ons and product (marke@ng) demos.
•
Our Mobile PlaPorm is controllable by Xbox controller over WiFi.
•
Break system SW into four major categories: Controller Interface, User Interface, Communica@on Interface, and Data, Modules, libraries [see Fig 2].
• • •
Real @me video streaming was obtainable with gstreamer library wriDen in C++.
Develop user interface with configurable seWng and live media streaming.
Speed Down
Speed Up
LiQ Down
Fig 9 Google SketchUp Model of the plenum crawl space.
LiQ Up
Incorporate easy-‐to-‐use scheme to control the robot.
Conclusions/Future Work
Fig 5 XBox Controller Layout
Fig 7 Graphical User Interface designed as a user-‐friendly workspace.
• •
JAVA Swing provides the API u@lized for the graphical interface and system status. Integrate Computer Vision Algorithms & post-‐processing capabili@es [see Fig 10]. o Open source SW, OpenCV and ImageJ, are used for automated detec@on and post-‐processing of images. Automa@c report genera@on – including all its en@@es. Add acous@c sensor to provide addi@on loca@on and system informa@on.
Crack DetecMon
Fig 2 High-‐level view of the flow and components structuring the System SW and enabling Communica@ons.
Fig 8 Google SketchUp modeling everything from the robot down to individual bolts with absolute precision. Acknowledgements: Electrical and Computer Engineering Department, Alex Irwin, Jack Golden, Dalton Colen, Kurt Braun
Object DetecMon
Hanger Straightness
Fig 10 Simulated Computer Vision Algorithms.
Rust/Corrosion detecMon
i93 Tunnel Inspection Robot
Joseph Robinson, Robert Watson, Sam Coe, Matt van Berlo, Josh Johnson Advisor: Professor Shafai Berham Abstract
State of the Art
Robot System
A mobile robo@c system has been built as a tool to improve the bridge inspec@on process done by the MassachuseDs Department of Transporta@on (MassDOT). Current methods for inspec@ng the plenums (ceiling space) are @me consuming, costly, and pose significant safety hazards to both the inspectors and the drivers. Our system will provide a means for inspectors to conduct these inspec@ons at the seat of their desks. On top of reducing cost, @me, and poten@al hazards, the mobile plaPorm enhances inspec@on reports with its soQware package offering automated & vision technologies.
• •
(b)
Inspec@ons are done while inspectors crawl in plenum place, and documen@ng notes and hand sketches are done this way for the en@re length [see Fig 3]. Lane closures are held for inspectors to enter and exit the plenum crawl space.
Head Light and Camera Flash
LiV Camera (Servo Controlled)
Background Star@ng back in 1991, the Big Dig remapped the 6-‐laned I-‐93 from above the city to below via a 3.5 mile long tunnel. Inten@on was to rid Boston of severe traffic concerns, which caused up to ten hours of traffic per day for city drivers. Projec@ons forecasted traffic reaching up to sixteen hours a day by 2010. In 2006, a 12t ceiling @le fell on a car and the drivers. AQer, several anchors embedded in the tunnel’s roof slab were discovered to be the cause. City-‐wide tunnel inspec@ons this problem as abundant, even in newly constructed areas [see Fig 1]. Fig 1 Showing the ceiling collapse (leQ) and simulated view of the cause (right).
MassDOT spends over $3M each year on plenum inspec@ons. Inspec@ons call for lane closures, reintroducing traffic conges@ons. Plus, working condi@ons have proven unsafe for both the inspectors and the drivers alike.
(a)
(b)
(c)
Fig 3 (a) A picture of the i93 plenum space inspected each year. (b) Entrance to Plenum crawl space. (c) Our envision of taking the inspector out of the tunnel and into his desk while performing inspec@ons.
LiV PlaXorm (3D Printed)
• Lane closures are held for inspectors to enter and exit the plenum crawl space. Base Camera w Bumper Protectors
Motor Controls
Results Fig 4 A close look at the prototyped Mobile System
Pause/Play Media
Camera 2 Control
• •
Mobile PlaPorm has been built for less than $1K [see Fig 4].
• •
Servos orientated parallel to ver@cal & horizontal results in 2D rota@onal control.
Launch Viewer
Camera 2 i/o
Set Flag/ Next Hanger
•
Record Video
Xbox controller provides an ease of use [see Fig 5].
Graphical User Interface with embedded media stream together with func@ons, features, and system informa@on [see Fig 7]. Google SketchUp model incorpora@ng accurate geometries and structural layout, along with the appropriate physics governing robot’s mo@on [see Fig 8 & 9].
Take Picture
Fig 6 Servos orthogonal to one another, providing rota@on in the ver@cal and horizontal. Base Control
Method
Base Control i/o Controller i/o
• • •
Provide MassDot a mobile plaPorm controllable from a PC with WiFi.
•
Build model for offline test simula@ons and product (marke@ng) demos.
•
Our Mobile PlaPorm is controllable by Xbox controller over WiFi.
•
Break system SW into four major categories: Controller Interface, User Interface, Communica@on Interface, and Data, Modules, libraries [see Fig 2].
• • •
Real @me video streaming was obtainable with gstreamer library wriDen in C++.
Develop user interface with configurable seWng and live media streaming.
Speed Down
Speed Up
LiQ Down
Fig 9 Google SketchUp Model of the plenum crawl space.
LiQ Up
Incorporate easy-‐to-‐use scheme to control the robot.
Conclusions/Future Work
Fig 5 XBox Controller Layout
Fig 7 Graphical User Interface designed as a user-‐friendly workspace.
• •
JAVA Swing provides the API u@lized for the graphical interface and system status. Integrate Computer Vision Algorithms & post-‐processing capabili@es [see Fig 10]. o Open source SW, OpenCV and ImageJ, are used for automated detec@on and post-‐processing of images. Automa@c report genera@on – including all its en@@es. Add acous@c sensor to provide addi@on loca@on and system informa@on.
Crack DetecMon
Fig 2 High-‐level view of the flow and components structuring the System SW and enabling Communica@ons.
Fig 8 Google SketchUp modeling everything from the robot down to individual bolts with absolute precision. Acknowledgements: Electrical and Computer Engineering Department, Alex Irwin, Jack Golden, Dalton Colen, Kurt Braun
Object DetecMon
Hanger Straightness
Fig 10 Simulated Computer Vision Algorithms.
Rust/Corrosion detecMon
