Latest news with #Auscape
News.com.au
23-05-2025
- Science
- News.com.au
Remarkable scenes as iconic landmark floods
Eroding river cliffs in the Lake Eyre Basin are a fascinating but often overlooked feature formed as a result of boom and bust hydrology, which refers to the extreme cycle of sudden flooding (boom) followed by long dry periods (bust). Some of the cliffs even reveal fossil layers or archaeological remains, offering rare insights into the basin's ancient history. Picture: Auscape/Universal Images Group via Getty Lake Eyre as seen from Space Shuttle Columbia during the Nasa's STS-35 Mission, December 1990. Picture: Space Frontiers/Getty Pink Moon by Luke Austin: A collection of aerial images of the iconic Lake Eyre. Picture: Ignacio Palacios/Supplied A group of award-winning photographers joined forces to capture the colourful pallets of the salt deposits, eroded channels and algal blooms of the uniquely Australian lake. Picture: Ignacio Palacios/Supplied Interpretations from the Air - Picture: Paul Hoelen/Supplied
Forbes
24-03-2025
- Business
- Forbes
Heat, Rain, Bumpy Roads & Hail, These LiDARs Cannot Fail
Dust hangs in the air after a mining explosion near Newman, Pilbara region, Western Australia (Photo ... More by: Auscape/Universal Images Group via Getty Images) The figure above shows the dust generated after explosives are used to loosen the ore so that the heavy machinery can extract and transport it for processing. Mining is dirty - and the environments are harsh and difficult for humans. Optical sensors go through these environments as they deliver critical perception, situational and localization to a computer that processes the data and controls autonomous movement of equipment and tools. The operations occur 24x7 in remote areas and availability of skilled manpower is scarce. Autonomy in mining environments is critical for productivity, efficient capital usage and safety. The challenge is for perception sensors (especially optical cameras and LiDAR) to perform under such harsh environments. Big machine autonomy is big business, as witnessed at the Consumer Electronics Show (CES 2025) earlier this year. It includes mining, agriculture, garbage disposal and airport tarmac operations, tasks which are performed 24x7 under temperature extremes, snow, rain, shock, vibration, dust and fog. Exhibitors at CES included Oshkosh, John Deere, Caterpillar and Komatsu, not necessarily consumer-product oriented, but certainly highly consumer-impacting. Their autonomy solutions address safety, productivity, 24x7 asset utilization, operational optimization and most importantly tackling the shortage of skilled human labor that must work in remote and physically challenging and harsh environments. Applications include tillage and seeding, excavation and transport of construction materials and mining ore, garbage hauling, sorting of recyclables, cargo handling in airports, and eventually autonomous construction on the moon! Figure 1: Planned Excavator for Lunar Construction Physical perception of the environment in which these autonomous machines operate is critical for safety (of people and equipment), navigation, path planning, speed control and optimizing operational protocols. Different sensor modalities are required, including cameras, LiDAR, radar, GPS, IMUs and gyros, along with software and AI-based fusion and decision making. The environments are harsh (compared to the on-road automotive use cases) - wide operating temperature ranges (- 40°C to 110°C), extreme shock and vibration, mud splatter, dust and mining ore contamination of sensors (which causes confusion in perception). Outer space operation adds another layers of environment challenges. Optical sensors (cameras and LiDAR) are especially impacted from a ruggedness and data quality perspective and require special hardware and software strategies to ensure operational integrity over long lifetimes. Caterpillar has been a leader in autonomy for mining, quarrying, oil exploration, and construction for the past 25 years. LiDAR is essential for obstacle detection since its ~1.3M lbs trucks move at 40 mph and long range detection of obstacles is critical for speed control, braking and steering decisions. Caterpillar started off using Velodyne LiDAR and later other suppliers like Leica and Ouster (Ouster acquired Velodyne 2 years ago) who supply products in the mining industry. Given the demanding operating environment, Caterpillar initiated internal LiDAR efforts in 2018 to overcome challenges with harsh environments (see Figure 2, especially the cameras in front of the cab section). Customized software and signal processing to filter false positives due to dust particles and deep ground ruts in oilfields is site-specific challenges. Caterpillar now works with commercial LiDAR companies under license to customize and adapt their COTs (Commercial-off-the-Shelf) products for different applications. Figure 2: The Cat Differential Steering System, designed in the 1980s, allows machines to turn 'on a ... More dime,' increasing productivity so customers can get more done in less time. Michael Murphy is the chief engineer in charge of mining autonomy at Caterpillar. As autonomy capability is tested in one situation (mining in remote regions in Australia), the challenge is to translate it into other locales and applications (like quarrying in semi-urban locations in the USA). Edge cases include dealing with iron ore dust in Australia and deep ruts in oil sands in Canada, confuse the perception stack and are addressed through signal processing and training. According to Mr. Murphy, Caterpillar is committed to providing our customers with reliable and durable solutions. LiDAR with fewer or no moving parts usually leads to increased component durability and reduced component maintenance. Doppler information is useful in some applications where there is a need to visually infer the speed of actors in the environment. It can also provide extra information useful for training AI models". The company is engaged in various verticals like farming, orchard spraying and lawn mowing tractors, and has invested in autonomy efforts for the past 25 years. A big driver for this is the scarce availability of trained manpower in the short time frames in which farms need to be tilled, seeded and harvested. The autonomy effort was accelerated with the acquisition of robotics company Blue River Technologies in 2017. Aaron Wells is in charge of developing autonomy capabilities.. Autonomous farming tractors are outfitted with 16 cabin-rooftop cameras (4 on each side, providing 3 unique stereo views in each direction. These provide good depth perception at a 20-30 m range in front and rear directions (important since tractors typically are dragging long attachments for different farming needs). Key concerns are addressing shock and vibration effects as well as durability. System calibration is critical (over temperature, lighting conditions, etc.) Insecticide-spraying in orchards presents a difficult perception challenge while maneuvering through tree branches (Figure 2). LiDAR plays a crucial role in mapping vegetation canopies and tree branches, as well as providing accurate localization and path planning for optimal coverage and efficiency. Figure 2: The Autonomous 5ML Orchard Tractor for Air Blast Spraying Uses LiDAR The approach is to pick a COTs LiDAR, and qualify and harden it for the harsh operating conditions. According to Mr. Wells, 'John Deere Autonomy teams are always looking to fit the right sensor to the needs of our application. Specifically for autonomy kits that contain lidar today, solid state lidar could provide benefits in ruggedness and form factor. Regardless of the type of lidar, some of our applications are less sensitive to the need for velocity tracking given the relative slow speed of these off road use cases. We're continually evaluating new sensors to meet these unique off road needs for our customers'. Founded in 1917 by two inventor-entrepreneurs, William Besserdich and Bernhard Mosling, to commercialize their disruptive 4-wheel drive designs for heavy vehicles, the company has grown to become a publicly listed corporation. It generates ~$10B in annual revenues, has 18,000 employees and 12 product brands serving markets in access, defense, fire & emergency, refuse collection, concrete placement and aviation ground support. Defense is a key segment (~25% of revenues) and the company invests significantly in R&D efforts for autonomy and safety. Oshkosh Defense is developing Leader-Follower technology in which a fleet of driverless autonomous vehicles follow a lead manned vehicle (Figure 4), reducing exposure to the dangers associated with battlefield movement, and providing increased flexibility in deployment. Autonomy developments like this are also extended to commercial businesses like refuse collection and aviation ground support (autonomous handling and transport of baggage, aligning passenger access jetways to airplane entrance doors). Figure 4: Oshkosh Autonomous Defense Vehicles Perform in Harsh Off-Road Environments The rationale for incorporating autonomy is to assist human operators, and improve safety and efficiency in physically challenging tasks, typically in outdoor environments. Perception is a key element to achieving this. Oshkosh leverages LiDAR, radar, cameras and GPS, with advanced machine learning algorithms for fusing sensor information to provide actionable situational awareness information capabilities for autonomy. For example, dust clouds are confused as obstacles by LiDAR, and if particularly dense, the sensor is blinded. Fusion with radar information alleviates this. Vibrations effects on camera images and LiDAR point clouds are filtered out by using IMU (Inertial Measurement Unit) and fiber-optic gyro data. Adapting such enhancements and edge case solutions to different environments is non-trivial and takes significant human effort. John Beck, Director for Autonomy and Active Safety, has been involved in the Deere's autonomy efforts for 2 decades. Harsh environments (shock, vibration, water and dirt splatter, dust, weather) are a challenge for perception and durability, especially for optical sensors like cameras and LiDAR. The approach is to select performance compatible COTs products and ruggedize them for different environments to achieve required lifetimes. . The point is that adapting these enhancements to different environments and applications is non-trivial and takes significant human effort. Per Mr. Beck, 'LiDAR with no moving parts is inherently more robust from a hardware perspective (moving parts tend to wear) and should be less expensive to manufacture and scale production. FMCW or 4D LIDAR using the Doppler effect to provide velocity of detections eliminates the need to calculate the velocity by measuring the returns from an object between 2 or more timestamps. This is highly desirable in some applications." Founded roughly a century ago in the town of Komatsu, Japan, the company today is a global provider of equipment for customers in forestry, construction, mining and quarrying verticals. Underwater dredging and lunar construction are emerging business areas. It is listed publicly on the Tokyo Stock Exchange and has revenues of ~$27B/year and ~65K employees worldwide. The mining business established its headquarters in the United States (Milwaukee, Wisconsin) after the acquisition of Joy Global in 2017. This enabled the integration of Komatsu's surface mining equipment with multiple brands of Joy's surface and underground products. The construction and forestry businesses are centralized in Japan. Wesley Taylor is the Senior Product Manager at the company's Autonomy Center of Excellence for Surface and Underground Mining. The Center is focused on integrating autonomy into Komatsu's suite of mining equipment. This includes hardware, software and physical simulation software of the equipment and processes. The environments encountered in mining are very different from those in automobiles. 24x7 operation is a given in locales ranging from the tar sands in Canada to the semi-desert environment in Western Australia. This requires the autonomy stacks to be tuned and calibrated specifically for a given customer, location and application. Figure 5 shows Komatsu's AHS system (completely autonomous without a driver cab), launched in 2021 and deployed on 750 vehicles to date, with more than 10B metric tons of material moves (at a rate of 6M metric tons a day). Figure 5: Komatsu Autonomous Trucks Operate Leveraging the FrontRunner Autonomous Haulage vehicle ... More (AHV) From a hardware perspective, the autonomy stack uses a combination of high precision GPS and IMUs, cameras, LiDAR and radar along with vehicular sensors monitoring tire and engine parameters. Sensor information is used with simulation models and machine learning to operate the vehicles safely and efficiently. For optical sensors like cameras and LiDAR, keeping the optical surfaces clean is critical and significant effort is expended on packaging and location of such sensors to prevent contamination from dust and ore particles. Sensor monitoring is used to recognize the presence of this contamination and deploy self-cleaning mechanisms. High precision IMU chips are co-packaged close to the LiDAR optical axis to enable filtering of the LiDAR point cloud due to vibrations (which are typically in the 50 Hz range). Komatsu works with its LiDAR suppliers to integrate IMUs in the LiDAR. Given the sensitivity of the LiDAR to performance to temperature, Komatsu engineered specialized thermal management solutions to address a wide operating range (-50 °C - +65 °C). From a LiDAR wish-list perspective, Mr. Taylor indicated that solid state is an attractive feature. Lack of moving parts minimizes shock and vibration, and fatigue-related lifetime failures (5-year lifetimes are a minimal requirement, higher is better). Large Field of View (FoV) is important given the size of the machines and mounting locations, and self-diagnostics to detect and report imminent failure of the LiDAR performance is critical. Doppler LiDAR is certainly attractive but not at the expense of range or point cloud density. The AoT™ (Autonomy of Things) revolution is progressing in areas that solve critical problems in harsh environments where labor is scarce, capital is expensive, and 24×7 operation is imperative. Advances in sensing, perception, localization, computing, and physical AI are making this revolution possible. Designing durable LiDAR sensors that are durable and provide accurate perception information under these environments is challenging and requires strong collaborations between LiDAR manufacturers and users.
