Researchers tested ShadowCam systems for autonomous vehicles in their offices with the wheelchair and in a parking garage with a car.
All photos courtesy of MIT
Changes in shadows can indicate something is moving before people or cameras can see anything. Massachusetts Institute of Technology (MIT) engineers have developed a system using that principle for sensing systems in autonomous vehicles.
In experiments with an autonomous car driving around a parking garage and an autonomous wheelchair navigating hallways, the shadow-sensing system beat traditional light detection and ranging (LiDAR) – which can only detect visible objects – by more than half a second.
“Where robots are moving around environments with other moving objects or people, our method can give the robot an early warning,” says Daniela Rus, director of the Computer Science and Artificial Intelligence Laboratory (CSAIL), the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science, and co-author of a paper on the development. “The big dream is to provide X-ray vision of sorts to vehicles moving fast on the streets.”
The system has only been tested indoors, where robots move slowly, and lighting is consistent.
MIT professors William Freeman and Antonio Torralba, who are not co-authors on the paper, collaborated on earlier versions of the ShadowCam system, which were presented at conferences in 2017 and 2018. It uses sequences of video frames to target a specific area, such as the floor in front of a corner, and detects changes in light intensity throughout time, from image to image. Some of those changes may be difficult to detect or invisible to the naked eye, but ShadowCam computes that information and classifies each image as containing a stationary or dynamic object.
A test wheelchair equipped with MIT’s ShadowCam identified when a person might be coming around a blind corner.
Adapting ShadowCam for autonomous vehicles required a few advances. A previous version relied on lining an area with augmented reality labels, similar to QR codes. Robots could scan labels to detect and compute their precise 3D positions and orientations.
To eliminate the tags, researchers combined image registration and visual odometry. Image registration essentially overlays multiple images to reveal variations. Medical image registration, for instance, overlaps medical scans to compare and analyze anatomical differences.
Visual odometry, used for Mars Rovers, estimates camera motion in real-time by analyzing pose and geometry in sequences of images. The direct sparse odometry (DSO) variant computes feature points in environments similar to those captured by the tags by plotting features on a 3D point cloud. A computer-vision pipeline then selects only the features in a region of interest.
ShadowCam uses DSO-image-registration to overlay all the images from the same viewpoint of the robot. Even as a robot is moving, it can zero in on a shadow to detect subtle deviations.
One test steered an autonomous wheelchair toward various hallway corners while humans turned the corner into the wheelchair’s path. With and without tags, ShadowCam achieved 70% classification accuracy.
Researchers also installed ShadowCam in an autonomous car in a parking garage. With the headlights off, ShadowCam detected the car turning around pillars about 0.72 seconds faster than LiDAR, accurately classifying images 86% of the time.
The high-capacity automatic tool changer (ATC) for BH series dual-drive press brakes increases production levels and utilization rates while reducing setup time.
The ATC allows for fully automated tool setup and removal, freeing operators to perform other tasks. Its 107.3ft x 170.6ft storage capacity provides tooling variation ranging from high-runner V-dies to specialized hemming or off-set tools.
Up to 2.0" V-die openings and 19.7" lengths can be placed with the single, large-capacity tool manipulator that has pick-and-place for small tool sections for setups not requiring full setup removal.
The BH series ATC machine can be used as a stand-alone press brake or automated system, enabling flexibility for one-off jobs.
GL double-edged inserts, available in 0.080" to 0.236" widths, are suitable for long-grooving applications. The insert is supported by two new geometries:
PR geometry is for medium-to-high feeds in various machining conditions. Its cutting-edge strength resists cracking when grooving interrupted cuts and is suitable for parting-off of bars.
PM geometry is for low-to-medium feeds on soft materials, such as austenitic stainless steel, offering greater resistance to built-up edge and reducing the bur left when parting-off tubes.
A line of universal tool blades, available in 1.02" to 1.26" heights, feature a double-sided clamping key and a design which enables inserts to be replaced with one hand.
ColorFuse integrates injection molding and painting for direct fabrication and surface coloring to quickly produce high-gloss and scratch-resistant parts and multi-textured exteriors with novel design options.
Material cost savings can be as high as 35% compared to wet painting while achieving the same high-gloss piano black appearance in an A-pillar application.
A molded part is flow-coated with a polyurea reactive lacquer while still in the multi-component tool and then leaves the mold in the desired color without extending the injection molding cycle.
ColorFuse metal-effect surfaces are scratch resistant and resilient to climatic and environmental effects such as UV radiation. Enabling precise accents and styling lines, ColorFuse offers high-gloss, matte, or grained surfaces.
The AMP MCP 18P monoblock connector is designed for harsh environments for industrial and commercial transportation with its flame-retardant composition and lever design.
For bulkhead, in-cabin, and fuse-and-relay-box applications, the unsealed monoblock connectors use the AMP MCP 2.8 contact system, have a mating force less than 75N, and can easily be used for new or existing designs. They fit into existing bulkhead cut-outs and the required secondary locking feature keeps contacts aligned.
Flame-retardant UL 94 V-0 material increases safety, as does the required secondary locking feature. In addition, four available colors ensure proper mating and ease assembly.
The TVS R2650 supercharger boosts performance in Ford’s 5.2L V8 engine that powers the 2020 Shelby GT500. The supercharger is the evolution of the Twin Vortices Series (TVS) platform, which features rotor coating for improved efficiency. The high-twist, four-lobe rotor design is 15% larger than the TVS 2300 supercharger found on the previous Shelby GT500 and features several improvements to maximize efficiency and improve performance at higher speeds.
TVS R2650 provides up to 12psi of boost, helping to produce 760hp and 625 lb-ft of torque in the Mustang. Technical modifications include a 170° helical twist of its rotors, which is 10° greater than previous TVS rotors. Other upgrades include bearing plate pressure relief points that reduce trapped volume pressure and optimized sealing for better flow efficiency.
Many orthopedic implant companies claim their technologies facilitate bone ingrowth (or allow bone growth). Little clinical evidence backs up these claims, other than titanium’s promotion of an environment for bone to grow. Plasma-spray or acid-etched titanium surfaces without substrate porosity, for example, may yield surfaces for initial expulsion resistance (preventing the cage from moving after implantation) and provide surfaces for bone to grow onto, but they yield little long-term ingrowth to provide the best fixation and long-term results.
Industry research has established common pore and strut sizes, however, further challenges must be studied:
Surface morphology of lattice structures – macro, micro topography
Lattice type, shape manipulation for optimal initial attachment, long-term proliferation
Correlation between optimal pore/strut sizes, macro stiffness of structures – preventing stress-shielding, adjacent biological issues
Manipulation of mechanical properties via computational methods (functional lattice grading, topology optimization)
Section view of a simulated vertebral body segment showing the ability to mimic the morphology of bone and its seamless transition from trabecular to cortical features.
Many software tools struggle to build the kinds of structures that will lead to the next generation of orthopedic implants. Furthermore, selective laser melting (SLM) machines are becoming more advanced, manufacturing complex structures with strut sizes smaller than 0.100mm (0.004") and features as small as 0.080mm (0.003").
Next-generation software provides tools to harness SLM capabilities, allowing advanced lattice-feature manipulation. Newly developed capabilities allow design engineers to take triply periodic minimal surface (TPMS) lattices and grade them to any ramped structure.
Rather than stop at linear remapping, any geometry representation can be altered in each axis with any complex parametric equation. Complicated models can also be generated via parameters or topology optimization and used to functionally grade structures.
Top, example structure showing the transition between a solid element and an internal latticed structure. Above, Final TPMS Mixed Lattice, with the location of a sphere controlling the gradual transition zone.
Using the inputs (See Figure 1), the volume lattice can be thickened with the Voronoi lattice input. The amount of thickening can be controlled precisely, given parameterization. Where the Voronoi lattice material exists, the final lattice thickness will be greater. The same principles can be applied when setting up the volume lattice – the cellular point spacing can be ramped via any representation that can be conceived and applied.
Complex lattices can also be transformed from one to another with no definable transition point, potentially limiting single weak shear points. Having the freedom to achieve the strut and pore size that produces biological benefits, and maneuvers the structural elements, allows for complex manipulation of orthotropic properties. This further helps engineers evaluate how bone ingrowth will change the macro stiffness of the construct and how to create lattice elements that provide long-term benefits for biological optimization.
Figure 2: Mixed TPMS structures, Lidinoid (L) to Gyroid (R) with different cell sizes.
Manipulating the mixed elements is not bound to a linear input. Objects can be mixed using complex structures or primitive models, (See Figure 2) where the same two TPMS lattices are mixed using an oblate spheroid. Where the sphere exists, the cells are a Gyroid TPMS structure, and the design moves from the sphere transitions to a Lidinoid cellular structure with no obvious transition point.
These tools transition complex lattice shapes to external solid features, simulating biological mimicry of bony structure (cortical and trabecular elements) and the seamless organic transition between them. This type of control over the internal and external structures, as well as discrete control of the transitioning elements, can help give implant developers the ability to alter structural properties on a macro and micro level. It also helps them understand how structures alter as bone integrates into them and engineer implants with specific ingrowth factors.
Figure 1: 1) A 20mm cube with non-variable, 10mm Gyroid TPMS cells. 2) The same 20mm cube with functionally graded Gyroid TPMS cells modified about 2 axes (X and Y). Note that the periodicity in the X and Y directions varies throughout the length. 3) The same 20mm cube with modified Gyroid TPMS cells, where the X-axis has been mapped with an X² function. 4) The first input is a regularly repeating volumetric lattice. 5) The second input is a stochastically generated Voronoi lattice at strut diameter ‘D’. 6) Selectively graded volume lattice using Voronoi lattice input as ramp feature.
This can be further refined by setting up transition phases from one structural element to another – to define continuous mechanical properties from entirely different structural types.
With more electric vehicles (EVs) on the road, manufacturers must ensure that battery packs will not leak under normal field conditions, creating a performance issue or safety hazard for drivers and passengers.
A leaking battery is more than just an inconvenience for the vehicle owner. Lithium-ion, the most common form of rechargeable battery for EVs, can burst into flame or even explode.
Leak testing these large and structurally complex packs poses unique challenges. While air-leak testing is well established, battery pack testing best practices are still evolving. How can manufacturers efficiently and cost-effectively ensure quality and assure consumers that they have nothing to fear when getting behind the wheel?
Size, stability
Manufacturers want to keep battery packs as light and cost-effective as possible by using enclosure materials such as aluminum alloys, glass-fiber-reinforced polymers, or thermoset vinyl hybrid resins. That means part stability may suffer, and common test methods can easily create destructive amounts of force. This issue is only compounded by the large size and internal volume of the pack. More volume means more surface area, which increases vari- ability due to factors outside the control of the leak test instrument.
This can make it difficult to simulate the exact leak conditions these packs will experience in the field while meeting production requirements.
An employee at Nissan’s Smyrna, Tennessee, Vehicle Assembly Plant builds a lithium-ion battery pack for a Leaf electric vehicle (EV). Testing battery packs for leaks is a critical safety step for automakers.
Expansion
Due to material instability, any test that uses air to build pressure inside the pack can cause the volume to expand like a balloon, increasing the measured leak rate. Parts that balloon during testing can present inconsistencies in repeatability due to elastic/plastic deformation of the part.
We are, as always, governed by the ideal gas law: pressure (P) x volume (V) = amount of gas in moles (n) x universal gas constant (R) x absolute temperature (T) of the gas or PV=nRT. We know that pressure, volume, and temperature are all related in a closed system (which a leak test is). If the part temperature changes during a test, the pressure within the part will change.
By the same token, the change in volume means change in measured leak rate. If the part changes in volume during the fill portion of the test, it is also susceptible to leak rate changes.
Temperature, pressure impact
Changes in environmental conditions, or testing for water intrusion, can lead to negative pressure within the pack. This can easily cause the pack to shrink, which can mask a potential leak.
Environmental changes are transient. They could be weather-related, but typically large effects are due to opening bay doors or running machinery. The more flexible the part, the larger the effect, meaning we must consider the pressure differential and decide which test approach – creating negative or positive pressure within the pack – is best.
Test type
It’s a long-running debate – which is better, a pressure decay leak test or a mass flow leak test? The answer, as always, is it depends. Pressure decay measurements are volume dependant, while mass flow is volume independent.
Mass flow does not require calibration factors or PD cals (the slope and offset to convert a pressure change to a leak rate) and will read the same leak no matter how large the part in question. It’s important to remember that the measured leak will still vary between parts if the measurement is taken before each part has stabilized and reached thermal equilibrium.
Pressure decay, on the other hand, relies on the calibration factor to give an accurate measurement. If you use a pressure decay test for flexible battery packs, leak rate must be based on a PD cal factor. Volume references for calculation of leak rate will result in incorrect results.
Mass flow tests do not require managing cal values or cal factors; therefore, it may be a better choice for large and flexible battery packs. The test method eliminates variables in measurement due to variations in part stiffness.
Now, to contradict myself, I have also tested flexible parts that were more consistent when using pressure decay. In these cases, the physical characteristics of the part, the plastic and elastic deformation it underwent during a normal test, responded better to allowing the pressure to drop during the stabilize and test phases. This halted the continual elastic deformation the part exhibited using mass flow for the given cycle time of the station.
An employee at Fiat Chrysler Automobiles (FCA) Windsor Assembly Plant in Canada installs the 16kWh lithium-ion battery in the Chrysler Pacifica Hybrid. The battery pack is neatly packaged under the second-row floor in a battery case that retains maximum interior volume for passengers and cargo.
Photo courtesy of FCA US LLC
Data time
Whether it’s pressure decay or mass flow, the next step is deciding how to ensure the most accurate, reliable test possible.
We do this by recording all the traceable waveforms from the test and correlating these with the related data for environmental effects. So, we can document and visualize the impact that any variable has on tests. By measuring and accounting for variables, we can improve process repeatability.
In other words, understanding and tracking activity in our test environment, and correlating this with what is happening within the pack, gives us insight to control and account for changes in the environment, to ensure the most repeatable and reproducible test result for an air-based test.
Conclusion
Better data provide better visibility, or as the old cliché goes, garbage in - garbage out. If you are not measuring the right things, you can’t account for the variables that negatively impact the measured leak rate. Achieving a reliable and repeatable leak test for EV batteries requires modern digital sensors and data analytics.
This allows you to track and measure the impact of external changes in the environment and provide the insight to account for them. And, it allows optimization of the test cycle to ensure the ideal result is achieved in the shortest time. This allows the test station to keep pace with the speed of production and avoid staffing an additional parallel test station or stations.
With the right data and means to analyze that data, you can find test limits faster, calibrate test setup faster, run simulations that allow you to immediately see the impact of changes to test parameters, and hit gage repeatability and reproducibility (Gage R) and cycle time targets.
