RightChain Routing Transcript

First page 1 Next page Last page

RightChain™ Routes Vehicle Routing Optimization Course Transcript

Edward H. Frazelle, Ph.D.

Dr. Frazelle (00:00): What's the difference between a route and a schedule? The route sequences stops and the schedule puts time to the seat. Those are very closely related. We're going to start with the routing piece and just by way of reminder, what is optimization? You write down the constraints. You write down the objective function and you find the solution that satisfies all the constraints and minimizes [00:00:30] or maximizes the objective function. Suppose we apply that to routing. There's an expression we use when we help companies set up metrics called what would mom do? Mom stands for mission optimization metrics. Write the mission statement for whatever the activity is. Then write the optimization statement and it will tell you [00:01:00] what you should measure. How would you know if you had done a good job setting up routes? What makes a good route? Every customer slash location is visited. Remember early on I said a three PL was relieved from their duties for saving their client too much money because they didn't meet all the pickups [00:01:30] and deliveries met all time window requirements. What's a time window? Dr. Frazelle (01:45): That's a time span. It's a period of time. The first of which I should not arrive before and the last of which I need to make sure I leave before. That's a time window. What else [00:02:00] minimizes fuel consumption Elapsed time meets hours of service requirements, meets safety requirements minimizes damage. [00:02:30] How could one route create more damage than another? Bad roads minimizes the probability of theft slash hijacking. If you're routing ships in today's world, better have this one on there. Maximizes [00:03:00] cube utilization. This sounded easy when we started. Anything else you want to add? Another one is balances workload. Dr. Frazelle (03:23): We did a project like this with Coca-Cola in Mexico. One of the big deals with them is the driver. Not [00:03:30] only do they deliver the goods, but they also collect the money and they're paid on commission. So if you don't give everyone the equal opportunity to earn a commission, you got some problems. Also, if you create these routes and one driver has six hours worth of work to do and the other has nine people get upset about things like that, right? So we're going to have to do some balancing in here. Anything else you want to add? Dr. Frazelle (04:00): [00:04:00] Very good. Meets traffic guidelines. Laws does not exceed the speed limit. Does not exceed weight limits on roads does not violate [00:04:30] one way. Streets minimizes wear and tear on equipment. How could one route do more damage to a piece of equipment than another? It's road conditions. Anything else you want to add? Sounded easy to start with. Routing. It's just a route. It's a lot harder than that now. That's the mission of routing [00:05:00] right there. How would you take those words and form them into an optimization? Let's go through these and see if we can characterize these as either objective function kinds of things or constraint kinds of things. Every customer is visited. Dr. Frazelle (05:21):

That's actually a constraint. This is not something you minimize or maximize. You meet a stipulated number of customers. You have [00:05:30] to visit that number of customers. That's a constraint. Meet all time window requirements. That's a constraint. You executed all the pickups and deliveries. That's a constraint. You're not trying to minimize or maximize that. That's something you have to do. Minimizes fuel consumption. Anytime you see that word right there, you should think that's an objective. [00:06:00] Minimizes the driver's time, minimizes the elapsed time meets hours of service requirements, meets safety requirements minimizes damage. That's an objective. Minimizes the probability of theft and hijacking. Dr. Frazelle (06:25): That's an objective. Maximizes the cube utilization. Balances [00:06:30] the workload. That's going to be a constraint. Meets the traffic guidelines, does not exceed speed limits. That's a constraint. It's not being adhered to, but does not exceed weight limits, does not violate one way. Streets minimizes wear and tear on equipment. I hope you're starting to pick up the difference. Now let's see if we can put that thing in the form of an optimization statement. I'm not doing this just [00:07:00] because we're at Georgia Tech. This is the hub of optimization in the whole world. What I have found is that if you look at ways to solve problems, especially in logistics, this is one of the best ways to solve it. We have an expression in our firm that says, don't philosophize optimize. Put it in this format, run the numbers with it and you usually wind up with a better solution. Let's try this. Let's make a table with constraints. [00:07:30] One was we visit every stop, make every pickup slash delivery. The hours for the route have to be less or equal to the hours that come from [00:08:00] government regulations. Now we're getting a little fancy less or equal to this is a good one. How could you write meets safety requirements in this form? You're going there. You're getting there. A, you might say that the probability of an accident is less than or equal to whatever target that you established. Dr. Frazelle (08:30): [00:08:30] You could also have the severity of the accident is less than or equal to some target. You can even multiply those things together that the probability of the accident times the severity of the accident is less than or equal to some target. That's how you start to convert these things [00:09:00] that are in laws and words into numbers. Can you calculate the probability of an accident on a route? Absolutely. And can you track the severity of accidents on a route? Absolutely. Okay. You want the probability of theft to be less than or equal to a target that's established speed limits. Dr. Frazelle (09:29): How can you enforce [00:09:30] speed limits in routing? You can literally force it with governors, with electronic speed regulators, and there's all kinds of things you can do there. I was driving up in Ohio one day this summer and there was a motorcycle doing wheelies in the middle of the interstate, so I called 9 1 1 and gave him the license plate and said, oh, you're not the first one who called? We have several officers on the way. That's how you start [00:10:00] to build a set of constraints. Now, let's try an objective function. One characteristic of a good objective function is the units of measure have to be the same. I can't say minimize hours plus, minimize gallons

plus minimize total elapsed time. You can't mix units of measure and objective function and you can't mix directions. I can't say maximize CU utilization plus minimize elapsed [00:10:30] hours. It has to be something you're trying to minimize or maximize in a common unit of measure. Otherwise, no software tool on the planet Earth can deal with it. No human being really can deal with it. That's the characteristic of an objective function. Same unit of measure and directly it's going one way. Here's what we had for objectives. Minimizes fuel consumption, minimizes the driver's time, minimizes the elapsed time [00:11:00] minimizes damage, maximizes CU utilization, minimizes wear and tear on the equipment. Is there some way to summarize all of those in a unit of measure? What are you really trying to do? Dr. Frazelle (11:22): Minimize cost. Who said that but you're trying to minimize [00:11:30] the cost of the route. Minimize total routing cost. Now what costs would be included in there? Fuel driver's time plus capital. The equipment plus maintenance, so if you take that objective [00:12:00] function with these constraints, that is an optimization statement for routing, but if you can take that through mission optimization metrics. Now given this, what should the metrics be for the performance of routing? What should I measure to figure [00:12:30] out how good a job I'm doing with routing given this cost per stop might be one. Total routing cost. What portion of that is fuel versus the driver's time versus capital versus maintenance? Dr. Frazelle (12:49): The hours each route is taking. How many accidents are occurring? The metrics are right there. Mission optimization metrics. How many trucks are in the Wawa [00:13:00] fleet and how many stores? How many ways are there to send 200 trucks to 580 stores? A giga trillion Coca-Cola, Luis, how many trucks in that fleet? A lot. Giga trillion Visiting a giga trillion locations. UPS the same way. Schneider I'm sure is the same way. These numbers are so huge that a human being is not going to be able to figure it out. They might [00:13:30] get started, they might get a good solution, but they're probably not going to get an optimal solution. What difference does it make in a big problem? It makes a big difference. Normally good routing software will be between five and 20% better than a human being. Now five to 20% on the total routing cost for the Wawa fleet, Coca-Cola fleet, the UPS fleet, [00:14:00] the Norfolk Southern Fleet, the Schneider fleet, these big fleets, those are big numbers. Dr. Frazelle (14:05): Many millions that you can afford to invest to make things better. In my PhD program, I took a course called Complexity theory. I wouldn't recommend it to anyone. The only thing we did was to figure out how difficult problems are to solve. We never solved one. That was the problem. To assign the problem to a classification [00:14:30] of difficult. This routing problem right here is in the most difficult class of problems there is, so if you ever wondered why is that so hard? It's in that classification that is so hard that if you just let a computer run forever and it was the fastest computer in the whole, it would never find the optimal solution. Dr. Frazelle (14:56):

Never. It's that difficult. Normally [00:15:00] what you have to have are some good, what are called heuristics. We're really nerding out now let's do some examples of heuristics here for just a little bit and you're going to get to try your hand at this today too. Let's suppose there's a distribution center here and these are the customers I've got to visit. I'm going to create a route. I want you to tell me if I'm doing a good job or a bad job. I do. [00:15:30] It was amazing. It was so good. It was amazing. Now, where did I go wrong? We started here. We went up here, we came down here, went over here, down here, back here, and stopped right there. Now, at some point along the way, you just went nuts. There was something I did that you really got upset about. What was that? Dr. Frazelle (15:53): Backtrack. In the absence of time constraints, a good route never crosses itself. [00:16:00] That's rule number one. We're going to build some rules for routing. These will be routing rules in the absence of time constraints, time, windows a good route never crosses itself. What else wrong with that picture? [00:16:30] We didn't visit all the locations, so that's going to be another one. We must visit all locations. What else is wrong? There's something really bad wrong. We didn't get back home finishes at the home location. [00:17:00] Let's try a few more. There's some simple rules you can use to help build good routes. Is that a good route? Relatively speaking, never crossed itself. We visited all the customers. What's missing? The dc This little algorithm is called Route first, [00:17:30] cluster second. Dr. Frazelle (17:36): Oftentimes what the software is doing is it creates a route between all the customers. Then based on the capacity of the vehicle because we hadn't even gotten into that yet, and the requirements of the truck, it'll create routes and then efficient routes and then it will try to interchange the start and stop points [00:18:00] and that's why you're watching the algorithm. You see the routes getting better and better and better and better, and if you let it run forever, which of course you can't do, eventually you get to the optimal solution, but you can't let it run for infinite time. That's why it's in this class of problems that is so difficult, but if you get a good route and it meets all these requirements, you're actually in pretty good shape. Now let's try this with some capacities. Here's the distribution center and let's say that our vehicle [00:18:30] can carry eight pallets at a time and they want four and they want five and they want one and they want two and they want six and they want three. They want 4, 1, 5, 8, 7, and two. Dr. Frazelle (18:48): Pure delivery, no pickups. The cluster first would say, let's find nodes that are close together [00:19:00] that are pretty close to this requirement right here. So somebody tell me where those might be. The first thing you look for is one all by itself. That's eight. That's a cluster all into itself, so I'm out and back. Where's another pair of two or three physically close together that make eight to six and two? [00:19:30] Here's two fours here and back. Is that a good set of routes? Did it meet the constraints? Yeah, we visited every customer with the required quantity. Now, could it be better? How would you figure out if it could be better?

Dr. Frazelle (19:57): You start interchanging, that's it. You start different seed points, [00:20:00] you try different clusters and you iterate that little heuristic. It's called cluster first, route second and interchange. And if you ever wondered what in the world's going on in the software, that's what's going on right there. The way the time windows work is you would have, let's say on here, you would have 3:00 PM to 5:00 PM 1:00 PM to 6:00 PM and that's just the next level of complication on here, so [00:20:30] the more advanced routing packages will have that included. There's some other tricks of the trade with routing and scheduling. These are routing principles and best practices. One is called pooling and consolidation. This is Polo Ralph. Lauren, what commodities of product does Ralph Lauren work with? Sweaters, shirts, pants, swimwear, [00:21:00] bedding. From a manufacturing standpoint, there are pockets of places that really get good at making a certain type of commodity like furniture. Dr. Frazelle (21:11): In High Point now it's actually in Indonesia carpet in Dalton, Georgia. The sourcing folks of the world have gone out and helped build up these bases of supply. Let's suppose sweaters are made in Sri Long Church from Korea pants in Malaysia, and these are all [00:21:30] in quantities that are less than a full container load. This is the ocean shipping and in the end, they all need to wind up in Long Beach. So would it be better to ship a portion of a full container load straight from Sri Lanka to Long Beach and one from Malaysia to Long Beach and one from Korea to Long Beach and one from the Philippines to Long Beach? [00:22:00] Or is it better to consolidate all of that merchandise into a full container load for the trip to Long Beach? Dr. Frazelle (22:14): Which is better? You got to run the numbers on it in general, depending on, lemme ask you, what does it depend on? It depends on the LCL rate from each of those ports [00:22:30] to Long Beach. It depends on how full the containers are. It depends on available capacity and the schedules out of those ports. In general. If you run those numbers, you'll find out it's a good idea. That's a consolidation program we're consolidating for the long trip across the ocean. So in general, if you've got small quantities produced near one another and they're going to make a long trip, you're better [00:23:00] off pulling those together for that distance. We did a program at Disney where they have lots of suppliers, small suppliers up in the northeast part of the United States shipping to Orlando. So instead of having each one ship LTL, we had a full truckload consolidation point and got a much better deal in the end. That's the idea of pooling and consolidation along those lines is another good practice in routing and [00:23:30] scheduling and that's the idea of backhauling. Backhauling goes like this. The idea of the principle is in the work you haul something on your way back. Let's suppose there's a load going full from the warehouse to a retail store door and it comes back empty. That last leg back is called deadheading. What is the utilization on that trip of the driver, the vehicle and the container? It's 50%.

Dr. Frazelle (24:00):

[00:24:00] Is there anything else that driver, vehicle and container could have done? Yes. Suppose there's a vendor that's going to ship something to the warehouse anyway on the way back, we might as well send the truck there and get, I'm just making the numbers easy here. 67% utilization of those assets instead of 50%. Is Backhauling always a good idea? [00:24:30] No. When would it be a bad idea? That leg right there is not free. This leg over here is not free, so you have to compare the incremental cost of this trip back with what the savings would be associated with 17% better utilization and that's a cost benefit analysis [00:25:00] that you have to do. The other thing that could happen is this takes time to make this stop and that trip you're going to have to go off the main route. If you need the asset back, it's not a good idea. Dr. Frazelle (25:13): So you have to understand the opportunity cost of a driver day, vehicle day and a container day [00:25:30] to know whether or not that's a good deal or not. Any of you have backhaul programs? Alright, at Johnsonville, what do you use for back haul? Anything you can get? Correct? Anybody else have a back haul program? David, great example. So at Honda you're going out full to the dealership and coming back with returns and cores, et cetera. Excellent example. [00:26:00] Anybody else? Yes. So Norfolk Southern, you're going out full with paper products and coming back with scrap paper basically Anybody else at Wawa? Frozen, grilled onions for cheese steaks only in Philadelphia. Louise Coca-Cola. So you go out full and then you [00:26:30] come back with containers. Cool. Anybody else? Yeah, Oris better not get that one on the tape Dr. Frazelle (26:46): Or as I do not know how the Iraqi army works. You know how the Iraqi army works? I don't know how they work. It's like a bus system for the troops. [00:27:00] That's really cool. 67% for whatever reason seems to be a rule of thumb out there for good utilization of these types of assets. I don't know why it worked out that way, but if you're there or better, you're usually doing pretty good. How do you find a backhaul? Opportunity Brokers. There's some websites that help. A lady in Atlanta runs a business. Her name's Ann Elliot and she helped people find backhauls. [00:27:30] I guess you could say she's a broker. I said, Anne, how do you do? She says, I just drive around the city and I see trucks that are empty and I think I wonder if there's something else they could be doing. Where are they coming from? And she just follows empty trucks around and keeps logs of backhaul opportunities and then calls people a couple of other routing and scheduling tricks of the trade and then we're going to put you to work. One is a practice called So far [00:28:00] what we have been talking about your fixed route, we set these things up at the beginning of the day and you pretty much know that's what you're going to be doing under continuous moves. You start with that, but then you get out here to a certain location and when you're done, you log into the system or you call somebody and you basically ask the system, what's the next best thing for me to do given [00:28:30] what I just did, how many hours I've been on the road and what else is happening in the network? And it runs those calculations along these same lines and it says, okay, what you need to do now is to go to Waco, Texas. You finished continuous moves. Dr. Frazelle (27:59):

that. Now what should I do? I only got two hours left in the day. Maybe I can't do anything or I don't know what's near Waco, Texas. It may find the closest thing to you within two hours or the furthest [00:29:00] thing from you. That is exactly two hours. Those are the types of algorithms that you put in there so you find the next best, the next best. And usually this will work, I would say between 10 and 20% better than just fixed routing. What's the problem with this? Dr. Frazelle (29:21): I don't know many people who want to deliver something in Waco and just say, okay, let's call the big [00:29:30] router in the sky and say, what's the next thing for me to do? Now you need to go to Secaucus, New Jersey. It's a lifestyle issue really, but honestly, there's some people who really that I wouldn't be one of them, but there are people who like that. Oh, I love the Oto SEC caucus. Let's see where the system tells me to go now, their husband and wife teams who travel all over the country doing this and it works great for them. Anytime you got a situation where congestion [00:30:00] is a big potential problem, you might want to simulate that activity at least once just to make sure the bottlenecks in the system based on the variability of what your activity is, are not going to put the system into gridlock. And so you can see where those bottlenecks are. That's the value of simulation. This is routing and continuous moves on steroids. Dr. Frazelle (30:28): This is called collaborative logistics [00:30:30] and it goes like this. Suppose there's this company called Georgia Pacific and they have a route that runs from Bangor to Buffalo and they have a route that runs Chicago to Green Bay to Chicago and out to Cedar Rapids. I don't want to drive that. That's cold. What is the utilization of the assets in that set of routes? We know it's 50% there. Here, it's 50% out there. You're just leaving the truck out there [00:31:00] and here it's a hundred percent. Now, unbeknownst to them, general Mills is running a route from Mechanicsburg, Pennsylvania up to Wells, Maine, then they're left. They run from Buffalo to Chicago and they're left and they run from Cedar Rapids to Mechanicsburg. What's the utilization there? Dr. Frazelle (31:26): It's not too good. Is it 50%? [00:31:30] If those two folks knew about one another, could they come together and make a better system? Absolutely. And it would look like this. So now we're from Chicago to Green Bay, back out to Cedar Rapids. We're full going this way. We're full going this way, we're empty. Right there we're okay. There we're okay there. What's the utilization of the assets [00:32:00] in that system? 90%, something like that. If you run the numbers on that for them, that was worth almost a million dollars. Is that a good idea? That's a good idea. How do you find out about those opportunities? You go network with the industry. You basically get out there and get involved in the industry to find out what these opportunities are and you think about it, who would be in that location that would need a load coming back the other way? You just sit down and if [00:32:30] you take a little while to brainstorm it, you'll identify the people pretty quickly if you just sit down and take the time.

Page 2 Page 3 Page 4 Page 5 Page 6 Page 7 Page 8

Made with FlippingBook - Online Brochure Maker