Truckl

#supplychain #transportation How autonomous are self-driving vehicles? The answer is layered in that there are multiple levels of problems and some require human intervention. Take Einride, the Swedish truck manufacturer, for example. Their vehicles are continuously monitored and support staff is ready to jump in when the trucks encounter an unforeseen situation. The vehicles drive in forestry and mining operations though, and there isn't that much that usually goes wrong. It isn't exactly the same as navigating a school zone in a densely populated urban area. As the NYT reports, robotaxis use remote assistants as well and it's important that they do. The complexity of where autonomous vehicles operate becomes apparent when we look at the environment in which they operate. One issue is that it changes constantly. When you train the algorithms on a specific street, for example, it will likely never be the same again. Different cars are parked along the road and traffic is certainly always different. It you relied on camera vision only this would be terribly confusing. There are obviously ways to alleviate the issue including always establishing a precise location. Then there are truly unforeseen situations such as new construction or an accident. This is where remote drivers come into the picture. They can look at the immediate environment of the vehicle to guide it past. Will we eventually see truly autonomous vehicles? I think so. However, it will be years before algorithms are capable of fully taking over. Until then drivers have changed from the seat behind a steering wheel to a seat behind a computer screen. #truckl #innovation

#supplychain #transportation Have you heard of "narrow AI?" It's basically the application of AI in very specific domains. Supply chain applications are a perfect example for this. An AI that analyzes historical data, customer behavior, market trends, and other factors can improve data forecasting, predict future demand, or optimize inventory levels. Route optimization is another where AI analyzes real-time traffic data, weather conditions, and historical delivery patterns to optimize delivery routes and avoid risks. Of course, companies such as OmniOpti have been doing this well before we had major trend towards AI. Other applications are predictive maintenance, supplier selection, quality control, automation of sorting and picking, or risk management. These narrow AI applications existed long before the term was coined. So, what is different now? First, the use of large language models (LLM) is truly new and allows for much more sophisticated research and analysis. It also ensures that you don't need an army of Ph.D.'s to actually run it. Asking a simple question will do. Second, the ability to compute has greatly increased. The GPUs we use today frankly leave prior processor generations in the dust. Third, and most important, we have access to more granular data furnished by a plethora of sensors and data capture points. This is key given that the quality of data determines the outcome. It is also a bottleneck in that you need to do far more work to obtain data than to process it. Models are easy, but capturing clean data from a hundred thousand endpoints is not. In fact, we won't see a barrage of disruptive applications for quite some time because of this. #truckl #innovation

#supplychain #transportation What are the benefits of reshoring manufacturing? I know this is a loaded question. A cynic might argue that it's not very feasible to contemplate this. In the short-term you focus on industries that are easily transferable, and in the long run on those that take more effort. To move away from one country and into another requires a lot more than just setting up a new factory, as you surely know. You need supplier networks that provide materials, a skilled workforce that populates your factory, and, most important, you need comparable unit cost to make it worth while. We can shift apparel manufacturing to France, for example, but labor will simply tripple the cost of each garment. Plus we'd be dealing with a strike every two weeks. It's more likely that we shift from one country to another. The good news for reglobalization, as this dynamic is now called, is that it spreads wealth to many other places. You might even argue that the shift away from today's dominant economies was in the cards all along. It just took a bit of a political push. As a country's standard of living rises, so will wages and expectations. When they do, your unit cost goes up and manufacturers begin to look for other spots on the planet to get things done. There are plenty of countries in SE Asia, Latin America and Africa that could greatly benefit, if we develop the infrastructure along with new economic opportunities. #truckl #innovation

#supplychain #transportation How hard is it for traditional manufacturers to move into software? You have surely heard the famous Marc Andreessen quote: "Software is eating the world." He said it in a 2011 WSJ interview and it illustrates why he is one of the most successful investors ever. Products have long transitioned into the intelligence business. We control our sous vide cookers on mobile apps and vehicles tell manufacturers about their products. This latter development really began in 1994 when the US mandated OBD-II, an on-board diagnostic system that records data. GM was one of the first to equip its vehicles with On-Star, which allowed drivers to talk to a concierge, and more importantly enabled the automaker to receive data. This is extremely helpful to establish how different components in the car perform, and naturally also how we handle those cars. Fast forward to today and you have Tesla which uses driver-facing cameras and steering wheel sensors to alert us when we get tired. Yet, traditional manufacturers are struggling to make the transition from assembly to in-car intelligence. VW is a great example. Several years back the carmaker hired over 10,000 software developers and then struggled to put them to good use. You can't manage software projects as if they were an assembly line after all. By now the company has all but given up on making the transition. Mind you, they still employ some 6,500 engineers through CARIAD and their internal teams is several thousand strong as well. Over the past six years, they spent $21.5 billion on acquisitions including taking a stake in Rivian, and at least 15 other EV, battery and software providers. If you can't beat 'em, buy 'em! #truckl #innovation

#supplychain #transportation Has TuSimple gotten away too easy last week? We've covered the autonomy developer on multiple occasions and they certainly looked very promising in the beginning. On the surface the settlement with federal investigators looks like a win for the government. The company and key executives will agree to no wrongdoing and pay $184 million. It is one of those weird legal situations, right? Why would you be paying many millions, when you have done nothing wrong? The saving-face argument is that people avoid a much larger lawsuit, and it eliminates the possibility of a wrongful conviction. Naturally courts are tough and defendants innocent. However, let's revisit what got TuSimple into trouble in the first place. The company went through a $1.35 billion IPO in April 2021 and did not use a SPAC (respect!). There were signs of trouble though. Its safety record had been beautified and the roadshow advertised autonomous trucking cost at 35 cents per mile. It's a surprise that this did not cause more of a stir. Why would the autonomy developer throw away a large part of its margin. There are also questions about how they could have possibly arrived at this number. Anyway, the more important concern is that Mo Chen, the founder of TuSimple, created a startup in China called Hydron. TuSimple's technology, which was built with funding from US investors, eventually somehow ended up in China. In this regard it sure looks like they got away pretty clean on all fronts: defrauding US investors and industrial espionage. #truckl #innovation

#supplychain #transportation Will the US ever have a large high-speed rail system? There are plenty of networks around the world. China's Shanghai Maglev travels at speeds of 460 km/h (286 mph) and the country has the largest network in the world with 45,000 km of tracks. The "bullet train" (Shinkansen) in Japan was the first high-speed rail system, beginning operations in 1964. The French TGV is one of the most well-known, with trains reaching speeds of 320 km/h (199 mph). France also holds the world speed record for conventional trains at 574.8 km/h (357 mph). German ICE trains operate at speeds of up to 330 km/h (205 mph), connecting major cities across Germany and neighboring countries. In the US, we have had Amtrak's Acela Express, which is technically not considered high-speed since it travels at 150 mph, not 155 mph which is the benchmark. However, several projects are making headway in the US. The existing Acela line in the Northeast will get an upgrade to now barely qualify for high-speed operation, although the trains only reach 160 mph and travel at much lower speeds for 2/3 of the journey due to track conditions. The most promising project is the Brightline West project between LA and Vegas, where construction broke ground in April. It cuts car travel between the two cities in half and will surely bring plenty of additional gamblers into the desert. Of course, there is an old project to connect Dallas to Houston, but it has been haunted by legal disputes over land, the organizers have done a fantastic job alienating public officials, and the necessary $30 billion in funding for the has not been fully secured. Maybe we'll just jump right into hyperloop on this one. #truckl #innovation

#supplychain #transportation How intelligent will our personal space eventually become? It is obviously clear that PCBs in coffee makers and fridges aren't the endgame. We will eventually live and work in spaces that include a plethora of sensors. One company that is developing technology in this area is MIT spinoff Butlr. The startup labels itself as a physical AI developer and builds sensors for senior living communities on the basis of body heat sensing and machine learning. The sensors magnetically attach to walls or the ceiling to provide insights into how a building’s interior is used for working and living. Data is captured anonymously and they do not collect personally identifiable information. This is a really interesting idea and the use cases go far beyond retirement homes, of course. Essentially, the company is building a platform to collect contextual data that is interpreted by algorithms to deeply understand what goes on. Butlr's position in senior communities is very smart given that they can do a tremendous amount of good, and there will be few barriers to adoption, which allows the company to build increasingly sophisticated solutions. What is more interesting though is that once the solution is in place, it can be applied to a plethora of problems. There are many useful applications in supply chain, for example, ranging from assembly operations to following material flows. It is also apparent that we will eventually see many sensors on trucks and trailers to monitor freight and conditions, especially when these assets operate autonomously. This is one area worth keeping a close eye on. #truckl #innovation

#supplychain #transportation Is there any benefit to going into a new technology early? If you work in the field, especially if you are selling innovation, you know all too well that almost everyone is "racing to be second." Let someone else try it first. Taking this stance is warranted in some situations, but it lacks one key element. If you understand an innovation, can build a business case, and you craft a vision for it, then you do not need to wait for anyone. This requires great confidence and deep emersion but yields large gains as well. There is a correlation between success and innovation. Companies that are regularly first in adoption tend to be very successful. In our industry Amazon is a great example. Last year the company invested $86 billion in research and development, which was just over 6.5% of revenue. The tangible result of these investments is that the company is the clear leader in several technology areas today. They literally started the warehouse robotics space when they bought Kiva, for example. What has happened in the twelve years since the Kiva acquisition is that Amazon has amassed an unparalleled trove of knowledge about warehouse robotics. Today, at a time when many people are scratching their heads about when to get into the game, Amazon is working on ways to better synchronize different robot types with one another. They are building algorithms that not only improve work safety and productivity, but deeply integrate a tremendous amount of data from backend systems into robotic processes. They could not any of this, had they not ignited this journey well over a decade ago. #truckl #innovation

#supplychain #transportation Wouldn't it be great if we could retrofit trucks to run autonomously? Just put a few lidar and cameras on your roof, connect a few cables and you're all set. Of course, if things were that easy someone would have done it years ago. The reality is that while you can put perception components such as lidar or cameras on any vehicle, they need to be tightly aligned and can't move around at all if you want accuracy. You have the issue that even when you understand the environment, you still need to affect the behavior of the vehicle. Implementing control components isn't easy either. There have been attempts to build platform-agnostic autonomy, but they aren't the norm due to the aforementioned issues. One company that develops vehicle-agnostic solutions is Polymath Robotics in California. The company focuses on industrial equipment and has created an autonomous tractor, called Farmonacci which, one might assume, is a clever play on the Italian mathematician Leonardo of Pisa, known as Fibonacci. Anyway, the approach is valid in that industrial equipment is easier to automate than road vehicles for obvious reasons. Polymath integrates into all major equipment sensors, which makes adoption fast and seamless at the end of the day. Users can build custom applications and leverage simulation to further develop solutions. All of this makes their approach highly flexible. It will be interesting to see whether this is the future for other autonomous equipment as well. #truckl #innovation

#supplychain #transportation Happy Labor Day! This holiday is one of those where it's easy to forget why we're celebrating. We honor and recognize the American labor movement and the contributions of workers to the country's development and achievements. Workers have built this country and workers are the ones keeping it together every day. According to the U.S. Bureau of Labor Statistics, there were 5,486 fatal work injuries in the United States in 2022, which is a sharp increase from the prior year. Work conditions aren't always a priority and it's important to remember that we benefit from the tireless efforts of those who are on the front lines: in warehouses, in factories, and on roads. In 2022, there were 1,115 combined fatalities among truck drivers and driver/sales workers. Of course, we should think about safety and working conditions every day, but today is as good as any to remember and to express our thanks to everyone who works hard. #truckl #innovation