Designing Under Constraints
Progress can't be saved in this browser, so this lesson's status can't be set.
Watch the video explainer
Learning goals
- Identify constraints, feasible designs, and an objective in a design problem
- Compare fencing layouts under a fixed budget and a changed site assumption
- Count the most items that fit within a specified packing layout family
- Justify a proven optimum over a stated set of possible dimensions
- Revise a packaging recommendation when an estimate of waste changes
Find the designs you can use
A constraint is a requirement a design must meet. For a garden, a spending limit and the available space are both constraints. A feasible design satisfies every constraint. The objective is what you want to make as large or as small as possible among the feasible designs. For the garden, it could be the area available to plant.
An assumption helps turn the real situation into something you can calculate. If you count the whole area inside the fence as plantable, for instance, you are assuming the model can ignore paths. When the design changes, keep track of which requirements and assumptions your calculation uses.
Worked example 1 Compare three garden plans
Problem. A school has a fencing budget of dollars for a rectangular garden. The two edges parallel to the sidewalk cost dollars per meter each. The two edges perpendicular to it cost dollars per meter each. Assume the site can fit all three proposals below, fence can be bought in the exact lengths needed, and the full enclosed area is plantable.
- Plan A. m parallel to the sidewalk by m perpendicular.
- Plan B. m parallel by m perpendicular.
- Plan C. m parallel by m perpendicular.
Find the cost and area of each plan. Which feasible proposal has the most planting area?
Solution. Call the dimension parallel to the sidewalk and the perpendicular dimension , both in meters. Multiply each dimension by its rate and by two because the fence has two edges of each kind.
For A, the cost is dollars, or dollars. Its area is square meters, or square meters.
For B, gives a cost of dollars. Its area is square meters, or square meters.
For C, gives a cost of dollars. Its area is square meters, or square meters.
| Plan | Dimensions (m) | Area (square m) | Cost (dollars) | Feasible? |
|---|---|---|---|---|
| A | by | Yes | ||
| B | by | Yes | ||
| C | by | No |
B gives the most planting area among the feasible proposals. C has more area, but its cost exceeds the cap. A leaves more money unspent, but the objective is planting area.
Check your understanding
A design costs exactly the budget, fits the site, and meets every other requirement. Its area is smaller than another feasible design's. How should you classify it?
The budget is a cap, so spending exactly that much is allowed. All requirements are met, making the design feasible.
The larger feasible design rules out an area optimum for the smaller one.
Recheck feasibility when the site changes
A best tested design wins among the feasible designs you actually compared. A proven optimum wins over every design allowed by the model. When the allowed choices really are a short list, checking every one can prove the optimum. If your list is a sample, you need another argument to cover the designs you left out.
Worked example 2 Use an existing fence
Problem. The school reconsiders the same three garden plans. It still has a dollar cap. Previously, each of the two edges parallel to the sidewalk cost dollars per meter and each of the two perpendicular edges cost dollars per meter. An existing fence can supply one whole parallel edge of any proposal at no cost. The other rates, dimensions, and planting areas stay the same.
- Plan A. m parallel by m perpendicular. Its old cost was dollars.
- Plan B. m parallel by m perpendicular. Its old cost was dollars.
- Plan C. m parallel by m perpendicular. Its old cost was dollars.
Find the new costs and choose the feasible proposal with the most planting area. What can you conclude if the three plans are only a sample?
Solution. Each plan saves dollars for every meter of one parallel edge. The savings are dollars for A and C and dollars for B.
| Plan | Area (square m) | Four new edges | One edge supplied |
|---|---|---|---|
| A | dollars, fits budget | dollars, fits budget | |
| B | dollars, fits budget | dollars, fits budget | |
| C | dollars, over budget | dollars, fits budget |
C now meets the budget. It is the best tested plan because its area is greater than the other two areas. Its area was already greatest of the three before the change. Using the existing fence made that area affordable.
If the list is a sample, the recommendation applies to these three plans. To claim a proven optimum, we would need to know the full range of layouts the site permits and show that none of them can do better.
Check your understanding
Four layouts are tested for a room. They all fit, and the fourth seats the most people. What extra fact would make the seating maximum a proven optimum from the table alone?
If the table includes every permitted layout, its largest feasible count is the maximum over that whole set. How the winner ranks on cost, floor area, or number of rows cannot establish that coverage.
Count a packing family
A layout family is a set of layouts that follow the same placement rules. Rules about rows, orientation, and clearance can turn a packing question into a count you can check completely.
Worked example 3 Pack tiles in two kinds of rows
Problem. A tray is cm wide and cm deep inside. Each rectangular tile is cm by cm. Use one layer with edges parallel to the tray edges. Each row runs across the tray width and has just one tile orientation. Rows may differ in orientation. Tiles can touch, and rows cannot overlap. What is the most tiles that fit under these rules?
Solution. A cm deep row holds tiles across. Three widths of cm use cm, and a fourth would need cm. An cm deep row holds tiles across. Five widths of cm use cm, and a sixth would need cm.
At most two cm deep rows can fit in cm. For each possible number, fill the remaining depth with as many cm deep rows as will fit.
| Number of 8 cm deep rows | Remaining depth (cm) | Number of 5 cm deep rows | Tiles |
|---|---|---|---|
The maximum in this layout family is tiles. One cm row and two cm rows use cm of depth and achieve that count. The cases cover all allowed row counts. For layouts that break these row rules, the table gives no upper bound. A different argument would be needed to decide whether any such layout can hold more.
Check your understanding
A tray is cm across. Each package in a particular row needs cm across. What maximum row count can you certify?
Three packages take cm, leaving cm. Four would take cm and exceed the row width.
Prove a minimum when dimensions can vary
When dimensions may be any positive real numbers that satisfy the constraints, a few test sizes can suggest a choice. To prove it is best, look for a bound that applies to every feasible size and check whether one feasible size reaches it.
Worked example 4 Find the least board for a divided box
Problem. A closed rectangular box with a lid needs of total inside volume and a height of cm. It has a full-height divider spanning the width. The box width can be any positive length up to cm, and there is no separate length limit. Ignore board thickness, overlaps, and cutting waste. Find the length and width that use the least board for the six outside faces and the divider. Prove the minimum and give the board area.
Solution. Measure the width and length in centimeters. The height and volume fix the base area at square centimeters.
The base and lid together need square centimeters of board. Two walls and the divider each have area . The other two walls each have area . Therefore the board area in square centimeters is
A board-area bound for every feasible width#
Try width cm and length cm. The two terms and have numerical values and . Their product is and their sum is . Compare that with two equal terms of : their product is also , but their sum is only .
For any feasible width, put and . Their product is always because is fixed at . The identity below compares their sum with their difference.
The square of the difference is at least zero. Since both terms are positive, their sum is at least . The bound is reached when the difference vanishes, so the two terms must both be . This requires width cm and length cm. The width meets the cm cap, and the volume is cubic centimeters, or cubic centimeters. The equality case is therefore feasible.
The minimum board area is square centimeters, or square centimeters. Every permitted size was covered because the proof used only the fixed product and positivity before checking the width cap at the equality case.
Check your understanding
A proof gives a lower bound of square centimeters for a box design, attained only at width cm. A revised rule allows widths up to cm. What does that proof still establish?
The equality width is no longer permitted. The bound still applies to the remaining designs, but a different argument must identify the best one among them.
Revise the cost calculation
Material and cost can rank two designs differently. A supplier might charge for cutting as well as for the board itself. And if a cut wastes material, the paid-for area can exceed the area that ends up in the box.
Worked example 5 Price the board that gets thrown away
Problem. A shop can order either of two closed rectangular packages. It needs exactly of capacity, and both designs fit the shelf. The board costs cents per square centimeter of board used or wasted. Each package also has a fixed cutting fee. Initially the shop budgets for no waste.
- Design A. Inside dimensions cm by cm by cm. Cutting fee cents.
- Design B. Inside dimensions cm by cm by cm. Cutting fee cents.
Ignore board thickness and overlaps. Which package costs less per unit at first? Which costs less if A needs an extra of board per package for waste and B still has no waste?
Solution. The products and both give , so both packages meet the capacity requirement.
For A, the six faces use square centimeters of board, or square centimeters.
For B, they use square centimeters, or square centimeters.
Multiply the purchased area by and add the cutting fee to get the cost in cents.
For A, gives cents initially.
For B, gives cents initially.
The revised estimate adds cents to A because the shop pays for more square centimeters.
| Design | Board in package (square cm) | Cost with no waste (cents) | Cost with revised waste (cents) |
|---|---|---|---|
| A | |||
| B |
A is cheaper at first by cents per package. With the revised waste estimate, B is cheaper by cents. An additional square centimeters for A would add exactly cents, enough for a tie. That threshold shows how sensitive the first decision is to the waste estimate.
Check your understanding
One package is cents cheaper than another. Only the cheaper design needs an extra square centimeters of board. At cents per square centimeter, what happens to the cost comparison?
The extra board adds cents, or cents, to the first design. That exactly removes its earlier cost advantage.